AI Response Evaluation
Assessing model outputs for accuracy, relevance, clarity, factual consistency, and instruction following using structured project guidelines.
Available for new opportunities
VINCENT CHIMAOBI
>
AI Evaluation Specialist · Prompt Engineer · Web Developer · Penetration Tester
I help evaluate AI systems, refine prompts, build structured digital workflows, and create professional web experiences with a focus on clarity, evidence, security, and long-term usability.
Belfast, Northern Ireland, United Kingdom
Belfast Time--:--:--
AI Evaluation SpecialistPrompt EngineerWeb DeveloperPenetration Tester
AI EvaluationFactuality VerificationHallucination DetectionRubric-Based ScoringResponse ComparisonEvidence-Based ReviewInstruction-Following ReviewOutput RankingError Pattern IdentificationModel Response AssessmentCriteria MappingPrompt EngineeringPrompt StructuringContext FramingConstraint DesignOutput Format DesignPrompt DebuggingIterative RefinementEvaluation Prompt DesignFew-Shot PromptingSystem Prompt PlanningWorkflow PromptingFailure DiagnosisPrompt OptimisationWeb DevelopmentResponsive Interface DesignComponent ArchitectureAccessibility ImplementationSEO OptimisationPerformance OptimisationStatic DeploymentInformation ArchitectureReusable ComponentsGitHub Pages DeploymentUI State ManagementWeb Content StructuringMaintainable Frontend SystemsPenetration TestingReconnaissanceVulnerability ValidationResponsible DisclosureEvidence CollectionImpact AnalysisReproduction StepsRemediation GuidanceWeb Application TestingRisk AssessmentSecurity Report WritingAnomaly DetectionScope-Aware TestingGenerative AI WorkflowsAI-Assisted StoryboardingVisual Consistency ReviewPrompt-to-Video PlanningAI Content StructuringNarrative Workflow DesignOutput Coherence ReviewScene RefinementMultimodal PromptingCreative Workflow AutomationAI Media ReviewData AnnotationLabel ConsistencyGuideline InterpretationHuman-in-the-Loop ReviewAnnotation AccuracyData Quality ReviewTask Instruction MappingOutput ClassificationEdge Case IdentificationAI EvaluationFactuality VerificationHallucination DetectionRubric-Based ScoringResponse ComparisonEvidence-Based ReviewInstruction-Following ReviewOutput RankingError Pattern IdentificationModel Response AssessmentCriteria MappingPrompt EngineeringPrompt StructuringContext FramingConstraint DesignOutput Format DesignPrompt DebuggingIterative RefinementEvaluation Prompt DesignFew-Shot PromptingSystem Prompt PlanningWorkflow PromptingFailure DiagnosisPrompt OptimisationWeb DevelopmentResponsive Interface DesignComponent ArchitectureAccessibility ImplementationSEO OptimisationPerformance OptimisationStatic DeploymentInformation ArchitectureReusable ComponentsGitHub Pages DeploymentUI State ManagementWeb Content StructuringMaintainable Frontend SystemsPenetration TestingReconnaissanceVulnerability ValidationResponsible DisclosureEvidence CollectionImpact AnalysisReproduction StepsRemediation GuidanceWeb Application TestingRisk AssessmentSecurity Report WritingAnomaly DetectionScope-Aware TestingGenerative AI WorkflowsAI-Assisted StoryboardingVisual Consistency ReviewPrompt-to-Video PlanningAI Content StructuringNarrative Workflow DesignOutput Coherence ReviewScene RefinementMultimodal PromptingCreative Workflow AutomationAI Media ReviewData AnnotationLabel ConsistencyGuideline InterpretationHuman-in-the-Loop ReviewAnnotation AccuracyData Quality ReviewTask Instruction MappingOutput ClassificationEdge Case Identification
In Focus
A Visual Tour of the Work
Seven perspectives on the disciplines, places, and mindset behind the work.
01 / 07
AI Evaluation
Structured, rubric-based review of model outputs.
By the Numbers
Measurable Impact, Built Over Time
0+
YouTube Subscribers
Grown through AI-assisted storytelling and video production.
0+
Total Views
Across AI-driven content channels built from scratch.
0+
Years in AI Evaluation
Across Remotasks and TELUS Digital, remote.
0
Core Disciplines
Evaluation, prompting, training, web, security, documentation.
About
A Disciplined Approach to AI, the Web, and Security
How evidence-based thinking connects evaluation, prompt design, development, and cybersecurity into one coherent practice.
Obasiochie Vincent Chimaobi is an experienced technology expert with a background spanning AI response evaluation, prompt engineering, data annotation, generative AI workflows, web development, and cybersecurity-informed analysis.
His evaluation mindset treats every project guideline as a judgement recipe — breaking instructions into clear requirements before rating outputs. He analyses each criterion by identifying what needs to be checked, what counts as full or partial fulfilment, and what evidence is required, then applies a structured five-step rubric approach to reach a final rating.
That same discipline carries into prompt engineering, where objectives become precise instructions with defined context, constraints, and output formats — and into web development, where responsive, accessible, and maintainable structure matters as much as visual polish. Cybersecurity work adds a risk-aware lens to everything: careful verification, error detection, and disciplined documentation.
Professional Strengths
Evidence-based judgementFast guideline adaptationClear written feedbackCareful error detectionRemote task discipline
Education
B.Sc. Physics Electronics
University of Port Harcourt
Rivers State, Nigeria
Graduated: 2021
Certifications
Udemy Prompt Engineering Certification
Junior Penetration Tester (eJPT)
Languages
English
Igbo
Location
Belfast, Northern Ireland, United Kingdom
Professional Experience
Penetration Tester & Bug Bounty Hunter
2025 — PresentIndependent Client Engagements / HackerOne / Bugcrowd
Remote
- Conduct penetration testing and security assessments, identifying, validating, and documenting vulnerabilities.
- Participate in responsible vulnerability disclosure programmes on HackerOne and Bugcrowd.
- Prepare technical findings with evidence, impact explanation, reproduction steps, and remediation-focused documentation.
- Bring cybersecurity-informed review skills into AI evaluation work — careful verification, error detection, and factual validation.
Generative AI Content Creator & AI Video Workflow Specialist
October 2024 — PresentIndependent AI Content Projects
Remote
- Built and managed AI-driven YouTube channels focused on AI-assisted storytelling and video production.
- Grew one channel to over 29,000 subscribers and more than 2 million views; another to more than 1.5 million views.
- Produced AI-generated Nollywood-style films using prompt engineering and structured creative workflows.
- Used ChatGPT, Claude, Gemini, Midjourney, Veo 3.1, and related platforms for research, prompt development, and content generation.
AI Evaluation Specialist
March 2022 — December 2023TELUS Digital
Remote
- Evaluated AI-generated responses using project guidelines, structured criteria, and quality standards.
- Assessed outputs for accuracy, relevance, clarity, factual consistency, instruction following, and overall usefulness.
- Reviewed responses to identify unsupported claims, incomplete answers, hallucination risks, and guideline violations.
- Applied careful, evidence-based judgement to support AI model quality improvement and response reliability.
AI Data Annotation Specialist
October 2020 — January 2022Remotasks
Remote
- Contributed to AI data annotation and evaluation workflows involving labelling, review, and quality control.
- Labelled and reviewed data against project-specific instructions, accuracy standards, and annotation guidelines.
- Evaluated task outputs for correctness, consistency, relevance, and compliance with required criteria.
- Developed a personal five-step rubric mindset to improve speed and consistency of judgement.
The Journey
From Physics to AI, Web & Security
A career built on analytical thinking, evidence-based judgement, and continuous learning — one milestone at a time.
2017
Started Physics Electronics Degree
University of Port Harcourt
Began a rigorous programme combining physics, electronics, and computational thinking — the foundation for a technical career.
Foundation in analytical & systems thinking
2020
First AI Data Annotation Role
Remotasks
Entered the AI industry through data annotation and quality control — learning how training data shapes model behaviour.
Entry into AI evaluation
2021
Graduated & Built Security System
University of Port Harcourt
Graduated with B.Sc. Physics Electronics. Final-year project: an underwater laser detection security system using Raspberry Pi and Python.
Degree + hardware security project
2022
AI Evaluation Specialist
TELUS Digital
Promoted to evaluating AI responses with structured rubrics — factuality, hallucination detection, instruction following, and comparative judgement.
Advanced to specialist evaluation
2024
Launched AI Content Channels
Independent Projects
Built AI-driven YouTube channels using prompt engineering and generative tools — reaching 29,000+ subscribers and 2M+ views.
29K+ subscribers · 2M+ views
2025
Active Bug Bounty Hunter
HackerOne & Bugcrowd
Began authorised penetration testing and responsible disclosure — bringing security-informed thinking into all technical work.
Active security researcher
Expertise
A Full Spectrum of Technical Capability
Twelve disciplines that work together — from AI evaluation and prompt engineering to web development and cybersecurity-aware thinking.
Prompt Engineering
Converting business objectives into clear, constraint-aware instructions that improve reliability and reduce ambiguity in AI workflows.
AI Training
Supporting machine-learning training workflows by preparing, labelling, and validating human-reviewed data with consistent quality.
Data Annotation & HITL Review
Human-in-the-loop labelling and review that keeps model behaviour aligned with project-specific accuracy standards and guidelines.
Web Development
Designing and building responsive, accessible websites with a focus on information architecture, performance, and maintainable structure.
Penetration Testing
Authorised security research, responsible disclosure, and risk-aware development thinking brought into every technical decision.
Technical Documentation
Clear, evidence-based written communication that makes complex systems, findings, and processes easy to understand and act upon.
Generative AI Workflows
Building end-to-end generative pipelines across text, image, and video for research, content, and storytelling use cases.
Analytical Thinking
Breaking problems into measurable parts, forming evidence-based judgements, and refining conclusions as new data arrives.
Hallucination Detection
Identifying unsupported claims, fabricated facts, and weak reasoning in model outputs before they reach end users.
Rubric-Based Evaluation
Designing and applying structured rubrics that turn subjective quality into repeatable, comparable, and auditable judgements.
Technical Problem Solving
Combining research, disciplined testing, and structured reasoning to resolve ambiguous technical challenges end to end.
How I Work
The Five-Step Rubric Mindset
A personal evaluation methodology developed over thousands of rated responses — structured, repeatable, and auditable.
- 01
Understand the Ideal Output
Before rating anything, I form a clear picture of what a fully correct response should look like — every constraint, format, and edge case considered.
- 02
Form an Initial Judgement
I read the response in full and form an instinctive impression. This initial read is useful, but never the final word — evidence must confirm it.
- 03
Map the Criteria
I break the guideline into discrete criteria, identifying what must be checked, what counts as full or partial fulfilment, and what evidence is required.
- 04
Compare the Evidence
I map the response back against each criterion, citing specific passages. Unsupported claims and hallucinations surface at this stage.
- 05
Finalise the Rating
I assign the rating based on guideline alignment, document the reasoning, and flag anything that needs a second review for consistency.
This approach treats every guideline as a judgement recipe — breaking instructions into clear requirements before rating, so each decision is defensible and consistent.
Proficiency
Depth of Experience, Visualised
A transparent view of where the deepest expertise lies — from AI evaluation and prompt engineering to web development and security research.
AI Response EvaluationAI Evaluation
92%Rubric-Based ScoringAI Evaluation
90%Hallucination DetectionAI Evaluation
88%Factuality VerificationAI Evaluation
87%Prompt EngineeringPrompt Engineering
89%Data Annotation & HITLAI Training
85%Generative AI WorkflowsGenerative AI
86%Technical DocumentationDocumentation
84%Web DevelopmentWeb Development
78%Penetration TestingPenetration Testing
75%Python ProgrammingDevelopment
72%Research & AnalysisAnalytical
83%Percentages reflect depth of hands-on experience, not formal certification scores. Updated as new skills develop.
Toolkit in Orbit
The Tools I Work With
A constellation of the platforms, languages, and security tools used across evaluation, prompting, development, and research.
AI Eval
CChatGPT
CClaude
GGemini
MMidjourney
VVeo 3.1
NNano Banana
PPython
NNext.js
RReact
RRaspberry Pi
HHackerOne
BBugcrowd
Evaluation & Prompting
ChatGPT, Claude, Gemini, Midjourney, Veo 3.1, and Nano Banana — used daily for response evaluation, prompt development, content generation, and AI-assisted video workflows.
Development & Hardware
Python, Next.js, React, and Raspberry Pi — covering web development, automation, and sensor-based security systems built from the ground up.
Security Research
HackerOne and Bugcrowd for responsible disclosure, with evidence-led methodology and remediation-focused documentation.
Hover the constellation to pause the orbit. Each tool is positioned by discipline — evaluation at the centre, generative platforms in the inner ring, development and security in the outer ring.
Full Stack
Every Tool, With Honest Proficiency
A transparent view of the full technology stack — filter by category and see exactly where the expertise lies.
C
ChatGPT
Daily use for evaluation, prompting, and content workflows.
Expert3 yrs
C
Claude
Complex reasoning tasks and long-context evaluation.
Expert2 yrs
G
Gemini
Multimodal evaluation and research assistance.
Advanced2 yrs
M
Midjourney
Generative image creation for content workflows.
Advanced2 yrs
V
Veo 3.1
AI video generation for storytelling content.
Advanced1 yr
N
Next.js
Production React framework with App Router.
Advanced2 yrs
R
React
Component architecture and state management.
Advanced2 yrs
T
TypeScript
Type-safe development across the stack.
Advanced2 yrs
T
Tailwind CSS
Utility-first styling for rapid, consistent UI.
Expert2 yrs
H
HTML5
Semantic, accessible markup foundations.
Expert4 yrs
N
Node.js
Server-side JavaScript for APIs and tooling.
Intermediate2 yrs
P
Python
Scripting, automation, and security tooling.
Intermediate3 yrs
P
Prisma ORM
Type-safe database access and migrations.
Intermediate1 yr
S
SQLite
Lightweight databases for static-friendly apps.
Advanced2 yrs
H
HackerOne
Responsible vulnerability disclosure platform.
Advanced1 yr
B
Bugcrowd
Crowdsourced security testing programmes.
Advanced1 yr
R
Raspberry Pi
Hardware security projects and sensor systems.
Intermediate3 yrs
G
Git
Version control and collaborative workflows.
Advanced4 yrs
W
WordPress
CMS for client websites and content management.
Advanced3 yrs
ExpertAdvancedIntermediateLearning
Principles
The Values Behind the Work
Six principles that shape every decision — from how an AI response is rated to how a vulnerability is documented.
Evidence Over Opinion
Every judgement — an AI rating, a security finding, a design decision — should be backed by evidence you can point to. Opinions are starting points, not conclusions.
Show the evidence, then the conclusion.
Clarity Is Kindness
Clear writing, clear structure, clear reasoning. Ambiguity wastes time and breeds mistakes. Whether in a prompt, a report, or a line of code — be understood the first time.
If it needs reading twice, rewrite it.
Security by Default
A risk-aware mindset belongs in every project, not just penetration tests. Small decisions compound into real safety — or real vulnerability.
Threat-model first, build second.
Document the Why
Code, prompts, and findings all outlive the moment they're created. Document the reasoning so the next person — or future you — understands not just what, but why.
Future you is a colleague. Be kind.
Right-Size the Solution
Not every problem needs a framework. Not every site needs a dashboard. Match the solution to the actual need — simple stays simple, complex earns its complexity.
The best tool is the one that fits.
Consistency Compounds
Showing up daily — evaluating, refining, documenting, learning — matters more than bursts of intensity. Small, steady work builds real expertise over time.
Discipline beats motivation.
AI Evaluation
Evaluating AI with Evidence, Not Preference
Structured, rubric-based review of model outputs: accuracy, factuality, instruction following, and hallucination risk assessed with care.
Instruction-Following Review
Checking whether a response actually does what the instruction asked — every constraint, format, and edge case accounted for.
Factuality Checking
Verifying claims against trusted sources and flagging statements that are unsupported, outdated, or internally inconsistent.
Hallucination Detection
Spotting fabricated facts, invented citations, and confident-but-wrong reasoning before they reach an end user.
Rubric-Based Evaluation
Applying structured rubrics so quality is measured consistently, compared fairly, and audited later.
Comparative Response Judgement
Reviewing multiple candidate outputs side by side and selecting the strongest based on evidence, not preference.
Human-in-the-Loop Judgement
Bringing careful, evidence-based human review into automated pipelines to keep model behaviour trustworthy.
Live Demo
What AI Evaluation Actually Looks Like
A real example of the kind of issues I detect in model outputs — factuality errors, hallucinations, and instruction violations, with the corrected response beside it.
PUser Prompt
Explain the difference between TCP and UDP and give one real-world example of each.
Model Response (Unevaluated)
5 issues foundTCP and UDP are both internet protocols. TCP is faster than UDP because it doesn't check for errors. UDP is used when you need reliability, like file downloads. TCP is used for video streaming and gaming.
TCP stands for Transmission Control Protocol and was invented by Google in 2015. UDP stands for Universal Data Protocol. Both operate at Layer 3 of the OSI model.
Detected Issues (5)
Factuality
“TCP is faster than UDP”Factually incorrect — TCP is slower due to handshake and acknowledgement overhead.
Instruction
“doesn't check for errors”Violates accuracy — TCP does check for errors via checksums and retransmission.
Hallucination
“invented by Google in 2015”Hallucinated fact — TCP was developed in the 1970s by Vint Cerf and Bob Kahn, not by Google.
Hallucination
“Universal Data Protocol”Hallucinated expansion — UDP stands for User Datagram Protocol.
Factuality
“Layer 3 of the OSI model”Incorrect — both operate at Layer 4 (Transport), not Layer 3 (Network).
Hover any highlighted snippet or issue card to see the connection. This is the kind of evidence-based, rubric-driven review applied to every AI output I evaluate.
Try the Rubric
Score a Response Yourself
This is the actual rubric used to evaluate AI responses. Click score levels for each criterion and watch the weighted total update in real time.
Instruction Following
25% weightDid the response address every part of the instruction?
Factuality
25% weightAre all claims accurate and supported?
Hallucination Check
20% weightAny fabricated facts or citations?
Clarity & Structure
15% weightIs the response clear and well-organised?
Completeness
15% weightDoes it fully answer the question?
Weighted Score
100/100
Excellent
Breakdown
Instruction Following4/4 · 25%
Factuality4/4 · 25%
Hallucination Check4/4 · 20%
Clarity & Structure4/4 · 15%
Completeness4/4 · 15%
Each criterion is weighted by importance. The final score reflects the weighted average — exactly how production AI evaluation works.
Evaluation Gallery
Common AI Mistakes I Catch
Real examples of the errors that slip into model outputs — and how evidence-based evaluation catches them. Click any card to see the issue and the fix.
Model Output
According to Smith et al. (2023), TCP was developed at Bell Labs in the 1980s.
The Issue
The cited paper does not exist. The model invented both the author and the publication.
How It's Caught
Verify every citation against a trusted source. Flag any reference that cannot be independently confirmed.
Model Output
Python was first released in 1995 by Guido van Rossum.
The Issue
Python was released in 1991, not 1995. The model stated an incorrect date with full confidence.
How It's Caught
Cross-check factual claims (dates, names, figures) against authoritative sources before accepting them.
Model Output
Here is a 500-word summary of the article: [proceeds to write 500 words]
The Issue
The instruction asked for a summary under 200 words. The model ignored the length constraint entirely.
How It's Caught
Check every constraint in the instruction against the output. Length, format, tone, and scope must all be verified.
Model Output
UDP is faster than TCP because UDP is faster. TCP is slower because it is slower than UDP.
The Issue
The explanation is circular — it restates the claim as the reason without providing any actual explanation.
How It's Caught
Identify circular logic by checking whether the 'reason' adds new information beyond the claim itself.
Model Output
Both TCP and UDP operate at Layer 3 (Network) of the OSI model.
The Issue
Both protocols operate at Layer 4 (Transport), not Layer 3. A fundamental networking error.
How It's Caught
Verify technical classifications against reference documentation — OSI layer assignments are well-defined.
Model Output
TCP is used for video streaming because it is faster, while UDP is used for file downloads because it is reliable.
The Issue
The entire comparison is reversed. TCP is reliable (for files), UDP is fast (for streaming).
How It's Caught
When a response describes trade-offs, verify that the characteristics match the correct entities.
These are illustrative examples of the kind of errors caught during evaluation. Each represents a real category of model failure that rubric-based review is designed to detect.
Prompt Engineering
Turning Objectives into Reliable Instructions
A practical discipline of defining context, constraints, and output formats, then refining prompts based on observed output failures.
Objectives into Instructions
Translating a business goal into a clear, testable instruction with defined context, constraints, and output format.
Context, Constraints & Formats
Specifying exactly what the model should know, what it must avoid, and the shape its answer should take.
Reliability & Ambiguity Reduction
Tightening language and structure so outputs become more predictable and less prone to drift.
Iterative Refinement
Refining prompts based on observed output failures — turning each failure into a sharper, more robust instruction.
AI-Assisted Workflows
Building reusable prompt components that support research, content, evaluation, and automation tasks.
Cross-Domain Application
Applying prompt discipline across web, research, content, and evaluation work for consistent quality.
Interactive
Try the Prompt Engineering Playground
Compose a structured prompt by selecting a role, task, context, constraint, and format. This is the same approach I use to build reliable, constraint-aware AI workflows.
Role
Who the model should beTask
What it should doContext
Who the audience isConstraint
Boundaries to respectFormat
Output shapeassembled-prompt.md
You are a domain expert with deep technical knowledge. Explain the following concept in detail with examples. The audience is a general, non-technical reader. Support every claim with evidence or examples. Format the output as a bullet-point list.
This is how structured, constraint-aware prompts are built — turning objectives into reliable instructions.
Need a custom prompt framework? Let's talkWeb Development
Building Responsive, Accessible, Maintainable Websites
A growing but credible technical direction — from lightweight static sites to dashboards, support systems, and client portals.
Right-Sized Scope
Not every website needs every advanced feature. Simple sites stay lightweight, static, affordable, and professional — while larger projects can grow into dashboards, content managers, or client portals.
Frontend First
Responsive layouts, accessible components, and clean information architecture built with React, Next.js, TypeScript, and Tailwind CSS for speed and maintainability.
Future-Ready
Static-friendly architecture that migrates cleanly to professional hosting when a project needs APIs, authentication, or a data layer.
Web Development Toolkit
Frontend & Interface
HTML5CSS3JavaScriptTypeScriptReactNext.jsTailwind CSSResponsive DesignAccessibility
Backend & Data Layer
Node.jsNext.js API RoutesPrisma ORMSQLite
Authentication & Automation
NextAuth.jsbcryptjsPython
SEO & Optimisation
SEOPerformance OptimisationStructured ContentMobile-Friendly Pages
CMS & Website Platforms
WordPressStatic WebsitesLanding PagesBusiness WebsitesPortfolio Websites
Advanced Client Features
Admin dashboardsContact formsFAQ chatbotsSupport inboxesMedia uploadsBlog managementBooking systemsNewsletter captureClient portalsRole-based accessWhatsApp integrationTestimonialsAnalyticsSEO-ready pages
Not every website requires every advanced feature. Simple projects can remain lightweight and affordable, while larger ones grow into dashboards, support systems, or client portals.
Cybersecurity
Responsible, Evidence-Led Security Practice
Authorised security research, responsible disclosure, and risk-aware development, documented with remediation in mind.
Authorised Security Research
Conducting assessments only within agreed scopes, with clear permissions and responsible engagement.
Responsible Disclosure
Reporting vulnerabilities through proper channels so they can be fixed before they are exposed.
Evidence Collection
Capturing clear, reproducible evidence that demonstrates impact without revealing unnecessary detail.
Impact Analysis
Assessing real-world risk — what an issue enables, who it affects, and how severe it would be if exploited.
Remediation-Focused Documentation
Writing findings that lead directly to fixes — clear steps, priorities, and verification guidance.
Secure Development Awareness
Bringing a security mindset into every build decision so safety is designed in, not bolted on.
Security Posture
Responsible Disclosure, Measured
A snapshot of security work — disclosure rates, response times, and evidence quality. All work is done under proper authorisation.
Current Threat Level
Low
No active threats. Routine monitoring.
Low15%
Moderate45%
High75%
Critical95%
Disclosure Metrics
Reports ResolvedGood
94%Disclosed vulnerabilities with confirmed remediation.
Avg Response TimeGood
18hAverage time to first response on security reports.
Responsible DisclosuresModerate
12Vulnerabilities disclosed through proper channels.
Evidence Quality ScoreGood
9/10Average evidence completeness rating on reports.
Metrics are illustrative aggregates. Specific vulnerability details remain confidential under responsible disclosure timelines and NDAs.
Projects
Work Frameworks & Expertise Areas
Conceptual project previews representing the kind of work delivered across AI evaluation, prompt engineering, web development, and penetration testing. Click any card for full details and demonstrations.
AI EvaluationVideoAI Evaluation Workflow System
A structured workflow for evaluating AI responses against rubrics, capturing evidence, and producing comparable quality scores across model versions.
Rubric-basedHITLEvidence Review
Prompt EngineeringVideoPrompt Refinement Framework
A repeatable framework for turning objectives into constraint-aware prompts, refining them based on observed output failures, and documenting what works.
Prompt DesignIterative RefinementDocumentation
Web DevelopmentVideoProfessional Portfolio Website
A responsive, accessible, maintainable personal portfolio built with a modern frontend stack — designed for static deployment and future migration.
Next.jsTypeScriptResponsiveAccessible
Web DevelopmentBusiness Website Planning Framework
A planning framework that helps small businesses choose the right scope — from lightweight static sites to dashboards, support systems, and client portals.
PlanningInformation ArchitectureScoping
Penetration TestingPenetration Testing Report Documentation
A documentation pattern for responsible vulnerability reporting — evidence, impact, reproduction steps, and remediation guidance without exposing sensitive details.
Responsible DisclosureDocumentationRemediation
Generative AIGenerative AI Content Workflow
An end-to-end workflow for producing AI-assisted video and storytelling content — from research and prompt development through review and publication.
Generative AIVideoStorytelling
Project visuals are conceptual demonstrations representing each expertise area. Click any card to view the full project breakdown and available animation.
Real Impact
Numbers With a Story Behind Them
Not vanity metrics — each figure represents genuine work, real growth, and measurable outcomes. Click any card to see the detail.
29K+
YouTube Subscribers
Grown organically through AI-assisted storytelling
Built a channel from zero using prompt engineering and generative AI video workflows. No paid promotion — all organic growth through content quality.
2M+
Total Views
Across AI-driven content channels
Multiple channels reaching audiences interested in AI-generated films, documentaries, and reenactment content.
4+
Years in AI
Evaluation & annotation experience
Continuous work across Remotasks and TELUS Digital, evaluating thousands of AI responses against structured rubrics.
12
Responsible Disclosures
Vulnerabilities reported through proper channels
Security findings documented with evidence, impact analysis, and remediation guidance — submitted via HackerOne and Bugcrowd.
The Difference
Why Work With Me
A clear comparison of what makes this practice different — not just in skills, but in how the work is done.
Aspect
Typical Freelancer
VCVincent's Approach
Communication
Slow responses, vague updates, missing context.
Fast WhatsApp replies, written progress updates, clear scope documents.
AI Evaluation
Gut-feel ratings, no rubric, inconsistent quality.
Evidence-based rubric scoring, documented reasoning, auditable ratings.
Security Thinking
Security as an afterthought, bolted on late.
Risk-aware from day one, threat-modelled approach, responsible disclosure mindset.
Documentation
Code with no comments, reports with no evidence.
Everything documented — the what, the why, and the evidence behind it.
Scope & Pricing
Vague quotes that creep upward, surprise costs.
Fixed quotes with clear deliverables, no surprises, right-sized scope.
Post-Delivery
Radio silence after handover, paid support for everything.
30 days included support, questions answered, small tweaks at no extra cost.
Not just deliverables — a fundamentally different way of working. Evidence-based, security-aware, consistently documented, and genuinely supportive after the work is done.
Activity
Consistency, Visualised
A year of professional engagement across AI evaluation, prompt engineering, web development, and security research — showing up regularly matters.
0
Active Days
0 days
Longest Streak
0 days
Current Streak
JanFebMarAprMayJunJulAugSepOctNovDecspacer
MonWedFri
Less
More
Illustrative activity pattern showing consistent engagement across evaluation, development, and security work. Each square represents a day; darker squares indicate more professional activity.
A Day in the Life
How a Typical Working Day Flows
A look at the rhythm of a day — balancing evaluation, development, security, and content work with discipline and focus.
AI EvaluationWeb DevelopmentCybersecurityContent CreationBreak
- 07:00Break
Morning Review
Review overnight evaluation queues, prioritise tasks, and check security advisories from disclosed programmes.
- 08:30AI Evaluation
AI Evaluation Block
Focused rubric-based evaluation of model responses — factuality checks, hallucination detection, and instruction-following review.
- 11:00AI Evaluation
Prompt Engineering
Refine prompt templates based on observed output failures, document patterns, and build reusable workflow components.
- 12:30Break
Lunch & Learning
Break for lunch while reading research papers or documentation on LLM evaluation and security topics.
- 13:30Web Development
Web Development
Build and review responsive, accessible websites — frontend work, component architecture, and performance optimisation.
- 16:00Cybersecurity
Security Research
Authorised penetration testing, vulnerability documentation, and responsible disclosure work on HackerOne and Bugcrowd.
- 18:00Content Creation
Content & Wrap-Up
Manage AI-driven content channels, review video outputs, and document the day's findings before signing off.
A flexible framework, not a rigid schedule — the exact rhythm shifts with project deadlines and client priorities, but the balance of evaluation, building, and security work stays consistent.
Website Concepts
Customizable Starting Points for Your Website
Professional website concepts that can be adapted to your brand, colours, content, and business requirements. These are starting points, not completed client projects.
Concept Preview
Small-Business Website
A professional website for a local or small business, with essential pages, contact information, and a clean design that builds trust with customers.
Customizable
Brand identityColoursServicesContact detailsImages
Concept Preview
Start-Up Website
A modern, energetic website for a start-up, designed to communicate vision, attract early customers, and showcase the product or service.
Customizable
Brand identityColoursProduct showcaseTeam sectionInvestor info
Concept Preview
Service-Business Website
A website for a service-based business, with clear service descriptions, booking or enquiry options, and trust-building testimonials or case studies.
Customizable
Service listPricing displayBooking systemTestimonialsContact method
Concept Preview
Company & Corporate Website
A professional corporate website with multiple departments, team profiles, company information, and structured navigation for larger organisations.
Customizable
Company brandingDepartment structureTeam profilesNews sectionInvestor relations
Concept Preview
Personal & Professional Portfolio
A personal portfolio website for a professional, freelancer, or creative, showcasing skills, experience, projects, and contact information.
Customizable
Personal brandingProject showcaseSkills displayResume sectionContact form
Concept Preview
Landing Page
A focused single-page website designed to convert visitors into leads or customers, with a clear call to action and minimal distractions.
Customizable
Headline and copyCall to actionLead capture formSocial proofColour scheme
Preview Coming Soon
E-Commerce Concept
A concept for an online store, with product listings, shopping cart, checkout flow, and product search. Available as part of the Advanced package scope.
Customizable
Product cataloguePayment integrationShipping optionsStore brandingCategory structure
Preview Coming Soon
Restaurant & Cafe Website
A website for a restaurant, cafe, or food business, with menu display, opening hours, location map, and online reservation or ordering options.
Customizable
Menu designPhoto galleryReservation systemLocation mapOpening hours
See a concept you like? It can be customized to match your brand, colours, and business needs.
Tell Me About Your WebsiteConcept designs are professional starting points, not completed client projects. Each concept is adapted to the client's specific requirements before development begins.
Website Packages
Clear Packages, Right-Sized Scope
Three cumulative tiers for your website project. Each includes everything in the tier below, plus more. Click any package for full details and feature explanations.
Starter
£250
A professional entry-level website for customers who need a relatively simple online presence.
9 features included
View full details
Popular
Normal
£450
A more complete website for growing businesses that need stronger content, visibility, and business integrations.
10 features included
View full details
Advanced
£650
A more sophisticated website for customers who need broader functionality, stronger integrations, or advanced capabilities.
10 features included
View full details
No active promotion right now. Join the priority list to get notified about the next website offer.
Prices are fixed public prices. Every project is quoted individually based on requirements. Third-party services, hosting, or subscriptions may incur separate costs, which are always clearly disclosed.
Estimator
Scope Your Project, See the Price
Select the features you need and get an instant indicative estimate. Every project is quoted individually, but this gives you a realistic starting point.
Core
Essential foundationsFeatures
Common additionsAdvanced
Complex systemsYour Estimate
Estimated range
£440–£570
Base (5 pages)£250
Responsive Design+£80
SEO Setup+£60
Contact Form+£50
Subtotal£440
Get a precise quoteIndicative estimate only. Final pricing depends on design complexity, content, timeline, and specific requirements. Cybersecurity work is scoped separately.
Credentials
Verified Experience & Certifications
A documented track record across AI evaluation, development, security research, and academic foundations.
Verified
AI / Prompt EngineeringPrompt Engineering Certification
Udemy2024
Comprehensive training on structured prompt design, context engineering, and iterative refinement for production AI workflows.
Verification available on request
Verified
AcademicB.Sc. Physics Electronics
University of Port Harcourt2021
Final-year project: underwater laser detection security system using Raspberry Pi and Python. Strong foundation in electronics and programming.
Verification available on request
Verified
Professional ExperienceAI Evaluation Specialist
TELUS Digital2022–2023
Evaluated AI-generated responses using structured rubrics, assessing accuracy, factuality, instruction following, and hallucination risk.
Verification available on request
Verified
Professional ExperienceData Annotation Specialist
Remotasks2020–2022
Contributed to AI training workflows through data labelling, quality control, and rubric-based output evaluation.
Verification available on request
Verified
Security ResearchBug Bounty Hunter
HackerOne & Bugcrowd2025–Present
Active participant in responsible vulnerability disclosure programmes. Publicly disclosable work includes Syfe.com and University of Port Harcourt.
Verification available on request
Verified
Generative AIGenerative AI Content Creator
Independent Projects2024–Present
Built AI-driven YouTube channels reaching 29,000+ subscribers and 2M+ views using prompt engineering and generative AI workflows.
Verification available on request
All credentials can be independently verified. Academic transcripts and certification IDs are available on request to serious enquiries.
Always Learning
What I'm Exploring Right Now
Technology moves fast. Here's what I'm actively studying to keep my skills sharp and current. Click any card to read the full learning breakdown.
78%
Advanced LLM Evaluation Techniques
Exploring multi-turn evaluation, adversarial prompting, and red-teaming methodologies for more robust model assessment.
Read full breakdown→
65%
Web Security & OWASP Top 10
Deepening knowledge of modern web vulnerabilities, secure coding patterns, and automated security scanning workflows.
Read full breakdown→
82%
Next.js 16 & Server Components
Mastering the latest React Server Components patterns, streaming, and partial prerendering for production apps.
Read full breakdown→
Writing
Notes on AI, the Web, and Security
Illustrative article previews reflecting the themes I write about as a freelance content writer. Real articles will replace these placeholders.
AI Evaluation
Why Rubric-Based Evaluation Beats Gut Feeling
A practical case for turning subjective AI quality judgements into structured, repeatable rubrics — and the five-step mindset that makes it work.
6 min readRead
DraftPrompt Engineering
Prompt Engineering as Documentation, Not Magic
How treating prompts like living documentation — with context, constraints, and clear output formats — produces more reliable AI workflows.
5 min readRead
DraftPenetration Testing
Security Thinking for Everyone Who Builds
Why a risk-aware mindset belongs in every project, not just penetration tests — and how responsible disclosure makes the web safer for all.
7 min readRead
DraftThese are illustrative article previews. Published writing will be linked here as it becomes available.
FAQ
Common Questions, Clear Answers
Practical answers to the questions collaborators ask most often.
Instruction-following review, factuality checking, hallucination detection, rubric-based scoring, and comparative response judgement. I work best on projects that value evidence-based, auditable ratings over speed alone.
Both. I build responsive, accessible sites from the ground up, and I can review and improve existing ones — from lightweight static pages to dashboards, support systems, and client portals. Scope is always right-sized to your needs.
Yes. I am based in Belfast, Northern Ireland, and I have spent years working remotely with distributed teams. Clear written communication and disciplined documentation keep everything on track.
Under Non-Disclosure Agreements where required. I document findings with evidence, impact, reproduction steps, and remediation guidance — without exposing sensitive details or unauthorised claims.
WhatsApp is fastest. Scan the QR code in the contact section or message +44 7882 753398. You can also email or connect on LinkedIn — I respond to all three.
Working Together
How a Collaboration Unfolds
From first message to final delivery and beyond — a clear, predictable process designed to remove uncertainty.
01
Initial Conversation
Day 1We discuss your goals, timeline, and scope on WhatsApp, email, or LinkedIn. I ask clarifying questions and give honest feedback on feasibility.
Deliverable
Clear understanding of requirements
02
Scope & Quote
Day 1–2I send a written scope document with deliverables, timeline, and fixed price. No surprises — you know exactly what you're getting and what it costs.
Deliverable
Fixed quote + project brief
03
Build & Review
Day 3–7+Work begins. For websites, you see progress early and can review direction. For AI evaluation, I apply the rubric methodology and document findings as I go.
Deliverable
Regular progress updates
04
Delivery & Handover
Final dayFinal delivery with documentation. For websites: deployment-ready code + handover notes. For evaluation: structured report with evidence and recommendations.
Deliverable
Final deliverable + docs
05
Support & Follow-Up
OngoingPost-delivery support is included. Questions, small tweaks, and clarifications are all part of the service — not an extra. Your success matters.
Deliverable
30 days included support
Ready to start?
The first step is just a message. Tell me what you're working on and we'll figure out the rest together.
Get Started
Build Your Website With Me
Tell me about your website project. I'll review your requirements and respond on WhatsApp with next steps. No account needed.
Already started a discussion? Continue your website enquiry on WhatsApp.
Refer & Earn
Referral Partner Programme
Know someone who needs a professional website? Refer them and earn 10% when they become a qualifying paying client.
10%
Commission on qualifying projects
How It Works
- 1Apply to become a Referral Partner
- 2Application is reviewed manually
- 3Approved partner receives a unique Referral Code and Link
- 4Share the code or link with potential clients
- 5Referred client submits a qualifying website enquiry
- 6Client becomes a qualifying paying client
- 7Referrer earns 10% of the qualifying service amount
Programme Rules
- Referred customer must be a genuine new client
- Self-referrals are not eligible
- The customer must identify the referrer through the code, link, or enquiry process
- Only one referrer can receive credit for a new client
- Commission becomes payable only after qualifying payment is received
- Cancelled or refunded projects do not qualify
- Commission is based on the eligible service amount actually received
- Third-party costs and externally purchased services are not automatically commissionable
- Promotions and additional discounts do not automatically stack unless explicitly approved
Example Referral Code & Link
Code:
VC-7Q2MLink:
website-url/?ref=VC-7Q2MCodes are assigned manually after approval. The website does not validate whether a code is officially approved.
Apply to Become a Referral Partner
Available year-round. Reviewed manually.
Stay in Touch
Occasional updates on AI evaluation insights, prompt engineering patterns, and new project work. No spam — just thoughtful notes.
Contact
Let's Build Something Reliable Together
Available for AI evaluation, prompt engineering, web development, cybersecurity, technical consulting, collaboration, freelance projects, and employment opportunities.
Available for new work
AI evaluation, prompt engineering, web development, cybersecurity, technical consulting, collaboration, freelance projects, and employment opportunities.
Location
Belfast, Northern Ireland, United Kingdom
