About Me

Two-person venture with four AI products live in production; I write the code, my co-founder runs business development from Dubai. Built Tablrr's two MCP servers from scratch rather than on top of a library, so external AI agents can run a restaurant directly — with a confirmation step before anything irreversible. Built Tripr's pipeline, routing each step to a different model and using a deliberately sceptical one to fact-check the others, plus a scheduler that fits a day around meal times and travel. Built ApplyTrack's acting AI chat across a web app and Chrome extension, and SkillPro for Universities — a white-labelled AI examiner that grades students against a taxonomy built from their professor's own course materials. Run it all on my own infrastructure — Docker, Kamal, Postgres, Redis, background queues, Sentry, GitHub Actions CI.

Built blockchain API integrations and CSV data mappers for cryptocurrency transaction importing. Designed ingestion pipelines to parse and classify transaction types (trades, transfers, staking, rewards) for automated tax reporting. Resolved data inconsistencies across third-party blockchain APIs and exchange file formats.

TopCashback — Built an AI-powered promotional content classification and rating system on Rails from initial prototype to production. Integrated RAG with vector embeddings to provide contextual data to LLMs. Designed similarity + consensus pre-filtering logic to keep LLM API costs scalable as volume grew.

SprintCV (Recruitment Platform) — Built AI-powered CV import with structured outputs across multiple LLM providers; candidate-to-job fit classification system; JD-tailored CV generator with multi-step pipeline; and a conversational AI assistant with function-calling tools for natural-language profile updates. Also delivered RESTful APIs for integrating SprintCV services with external platforms.