Open to AI product roles · Accra, Ghana · Remote

Henry Iddirisu · AI product engineer

I take LLM features from prompt to production UI, on platforms with real users and real money.

Four years shipping production systems in Python, TypeScript, and PHP: the backend behind an AI voice-agent platform, search-grounded Gemini pipelines, and real-time features for 20,000+ students.

LLM pipelines with evalsAI voice-agent platform backendEnd-to-end product featuresReal-time product at 20,000+ users

Proof at a glance

students on the real-time learning platform I ship to
20,000+
students on the real-time learning platform I ship to
shipping production systems in Go, Python, TypeScript, and PHP
4+ yrs
shipping production systems in Go, Python, TypeScript, and PHP
payment providers behind one checkout I built
5
payment providers behind one checkout I built
codebases per feature: API, tutor app, student app
3
codebases per feature: API, tutor app, student app
Recruiter? Paste your JD.See how my LLM pipeline scores the fit.

Positioning

Strongest when the AI feature, the product around it, and the deploy are one job.

AI with guardrails

A Claude support assistant that only answers above a confidence threshold, always offers a person, and stayed off until it passed an eval on real past conversations. AI essay suggestions tutors accept or reject. The JD Analyzer on this site, with evals checking its flags.

Features shipped end to end

A live lesson whiteboard across a Laravel API, WebSocket broadcasting, the tutor app, and the student app. An IELTS writing-corrections workspace the same way. I take the feature, not the ticket.

Platforms built from the first commit

Megatron, a supply-chain security platform in Go: repository scanning, five vulnerability feeds, signed CISA-ready SBOMs, and Veracode, SonarQube, ServiceNow, and Cortex integrations. Before it, the FastAPI backend for an AI voice-agent platform. Both built solo.

Production is part of the job

I owned GitLab CI/CD and Docker deploys for Mango Capital and fixed what only breaks in production: exhausted DB connections, CORS with credentials, migrations that fight live containers.

Selected work

Featured case studies and recent project work.

See projects
LaravelLaravel ReverbPHP

Lesson Whiteboard

Role: Full-stack engineer (sole developer across all three codebases)

A live whiteboard for online language lessons, built end to end across a Laravel API, Reverb WebSockets, the tutor app, and the st…

Writing

Notes from shipping AI features and production systems.

Browse writing