ABOUT
ME

• •Stepped in pre-launch while brands were already being onboarded, closing the gap between engineering output and the founding team's product vision to get the site and app launch-ready.
• •Kept engineering aligned with product intent by authoring PRDs and making UI/UX and feature calls across the launch.
• •Owned PR review, merge approval, and beta pushes — merged-PR throughput rose 122% (2.2→4.9/day) with median merge time down 71%.

• •Owned the first MVP phase of a healthcare IVR end-to-end — from scoping and user-journey mapping to AWS deployment.
• •Ran LLM rubric evaluation research on GPT-4 scoring consistency across temperature/top-p configs; turned findings into a PRD that got approved and adopted as a client proposal.
• •Reduced edge-case failure rate by 20% (from ~40 failing test cases) through systematic prompt engineering, guard railing, and structured system prompting across production agent flows.
• •Cut edge-case failure rate by 20% through prompt engineering, guardrailing, and structured system prompting across production agent flows.
• •Optimized STT → LLM → TTS pipelines across voice, SMS, and email — brought response latency from 4–5s to under 2s.

• •Evaluated LLM outputs for quality, accuracy, and alignment; fed structured feedback into model improvement pipelines.

• •Tested a mobile game and gave live feedback on mechanics, UI/UX, and retention — verbal sessions of 45–60 mins with the dev team.

Discovery

I start by writing down what the problem actually is before touching anything. Scope first — otherwise you're just building in the wrong direction fast.

Build

Once the scope is clear I get into the workflow. Prompt architecture, tool calling, fallback paths — figuring out how the pieces fit together before deployment.

Ship

Honestly I ship when it's good enough. Production breaks things in ways testing never does, and that's fine — that's where the real data is.

Iteration

Then it's both — fixing prompts based on what broke, and re-scoping features when real usage tells you the original assumption was wrong.

Product

PRD authoring · MVP scoping · Feature prioritization · OKR/KPI tracking · User-journey mapping · Competitive benchmarking · Cross functional team coordination · PR review · Product syncs · LLM evaluation & guardrail definition · Latency, cost & reliability tradeoffs

AI / LLM & Agents

LLM pipelines (GPT, Claude, Gemini) · RAG (FAISS, Sentence Transformers) · LangGraph · GraphRAG · ReAct agents · Agent frameworks · TTS solutions · Vector stores · Prompt engineering

Tools & Languages

Python · SQL · JavaScript · FastAPI · AWS Lambda/SES/SNS · Docker · AI-assisted coding · Google Antigravity · OpenAI Codex · Claude Code · GitHub Copilot · Make.com