Web · Live on Vercel
DoorGo
Property and hostel platform: bookings, rent payments, maintenance.
★ … ·
Welcome · --:-- in Nairobi
Open to remote work · 3+ yrs experience
AI Evaluation & Red Teaming Specialist
I design adversarial safety evaluations for frontier AI agents and keep annotation pipelines honest.
currently
Freelance AI Red Teamer
Alice Collective · Nov 2026 – Present
AI Annotator
Multimango (RWS) · 2026 – Present
AI evaluation specialist with hands-on experience in LLM red teaming, multilingual data annotation, and customer operations. Currently designing adversarial safety evaluations for frontier AI agents while supporting AI annotation workflows.
Background spanning AI, healthcare, logistics, and digital services — calm, precise, and at home in remote-first teams.
Day to day I split my time between freelance red teaming for Alice Collective, AI annotation with Multimango (RWS), and my computer science degree — so I see frontier-model safety from the builder's, the annotator's, and the student's angle at once.
Currently learning: advanced adversarial eval techniques for frontier agents

Taking on freelance AI evaluation and annotation work. Best reached by email — I reply within a day.
Design adversarial prompt-injection tasks used to post-train frontier models. Author YAML scenarios with personas, attacks, verifiers, and rubrics; validate each against state-of-the-art models.
Key win: iterated a Booking PIN-phish eval to 50% attack success across dev rollouts.
Review text-to-speech and conversational AI outputs for quality and guideline adherence across multilingual datasets.
Key win: consistent rubric application across multilingual TTS and chat datasets.
Coordinated delivery operations and dispatch; fixed bottlenecks from order data and drove a 15% lift in repeat orders within two months.
Key win: +15% repeat orders in two months from order-data fixes.
Scheduling, follow-ups, and insurance coordination with strict digital-record privacy compliance.
Key win: reliable scheduling and follow-ups with privacy-compliant records.
e-Citizen services and computer-basics training for walk-in customers.
Key win: trained walk-in customers in computer basics and e-Citizen services.
Programming, information systems, and database management — building toward AI systems and data.
Data cleaning, analysis, and visualization with spreadsheets, SQL, Tableau, and R.
Computer packages, internet research, and digital productivity tools.
Graduated with a B plain — the foundation for university computer science.
End-to-end red-team tasks: attack narrative, seeded personas and context, success criteria — iterated against frontier models until they bite.
Production YAML scenarios with entities, plant data, deterministic verifiers, and model-graded rubrics on the AgenticVerse platform.
Review of text-to-speech and conversational AI outputs against rubrics — consistent labels and documented edge cases.
Spreadsheet, SQL, Tableau, and R workflows: cleaning, analysis, and charts that answer the question. Google Data Analytics certified.
Quality and guideline-adherence review across multilingual conversational datasets for model improvement.
Continuously engaged across study and work since 2022.
Featured project
Hostels, bookings, and rent payments in production.
Web · Live on Vercel
Property and hostel platform: bookings, rent payments, maintenance.
★ … ·
App · TypeScript
Story-sharing app with mobile, web, and backend releases.
★ … ·
Web · React + Firebase
Dental clinic website, live in production.
Independent client build · code on request
Available as a professional referee.
Available as a professional referee.
Available as a professional referee.



Press kit available on request — just ask.
Red teaming · 3 min
My loop: map the surface, plant the trap, measure the bite.
Read postAnnotation · 3 min
What labeling thousands of outputs taught me about breaking models.
Read postCraft · 2 min
Every text item machine-checked inside the page frame.
Read postPrefer email — fastest way to reach me.
Replies within a day
Proposal template · Process one-pager
Two revision rounds included on every engagement.
Usually within a week for annotation and eval work; larger builds depend on scope — ask and I'll give you a date.
Short written brief, milestones with demos, async updates on your channel of choice. See the process one-pager for the full flow.
Yes — pick "Book a call" in the form above and suggest two times (EAT). I confirm within a day.
Exclusively — Nairobi-based, remote-first, comfortable across time zones.