AI reliability & LLM-evaluation engineer — RAG, agent tooling, and AI workflow automation — who also ships full-stack.

I build the automation, then I build the check that keeps it honest.

Most people stop when the AI works. I treat that as halfway. I build LLM systems end to end and go deep where it matters most: evaluation harnesses (LLM-as-judge, structured-output and graph-based metrics) wired into CI as release gates, retrieval pipelines over vector databases, and multi-provider agents — then I pair every automation with an eval layer so output is correct, observable, and defensible, not just a demo.

🔗 Portfolio: olamidotun-loto-portfolio.vercel.app

• AI reliability & evaluation — DeepEval + LLM-as-judge against curated golden datasets, faithfulness / groundedness / hallucination checks, and regression and eval gates in CI. The check is the part most people skip.
• AI / LLM engineering — RAG pipelines with hybrid (vector + keyword) search, multi-provider agents (Anthropic, OpenAI, Google Gemini), MCP servers, and PydanticAI agent workflows.
• AI workflow & automation — multi-step LLM/agent pipelines and no-code orchestration (Make, Zapier): event-driven flows, webhooks, queue-backed processing, and content pipelines with a stronger-model vetting gate so automated output is checked before it auto-publishes.
• Full-stack product — TypeScript across the stack: React / Next.js front ends, Node and Python (FastAPI) services, Postgres/Supabase with row-level-security multi-tenancy, real-time (WebRTC, websockets), embeddable widgets, and edge logic on Cloudflare Workers.

• 🎚️ audio-lab — a browser Digital Audio Workstation: a multi-track timeline on a Web Audio engine, a Python DSP backend, and natural-language AI control of effects through structured LLM output.
• 🎥 just-us — a real-time two-person video experience built on WebRTC, with synchronized in-call mini-games, scheduling, and invite flows.
• 🌳 family-tree — a multi-tenant family-tree platform with AI portrait generation, voice-to-text profiles, and a relationship-inference engine.
• 🧪 images-comparison — a multi-model face-recognition tool (DeepFace / InsightFace) that cross-checks several models against the same input instead of trusting one.

TypeScript · Python · React · Next.js · Node · FastAPI · Supabase / Postgres · WebRTC · Web Audio · Weaviate · pgvector · Anthropic · OpenAI · Gemini · Cloudflare Workers · Tailwind · Docker

RAG · LLM-as-judge · DeepEval · golden datasets / eval gates · multi-provider routing · agent orchestration · MCP · PydanticAI · Make · Zapier · webhooks / event-driven · human-in-the-loop vetting

⚡ Currently focused on AI reliability & evaluation, RAG, agent tooling, and eval-gated automation.

📫 Open to conversations about AI reliability / LLM-evaluation and full-stack AI roles — olamidotun-loto-portfolio.vercel.app · lolamidotun@gmail.com