AI-Focused Full-Stack Engineer
I build production AI, RAG over vector search, and multimodal document understanding. AI-focused full-stack engineer. I ship real LLM systems, RAG chatbots backed by vector search, and multimodal (VLM) pipelines that pull structured data out of messy, multilingual documents, and own them end to end, from the retrieval and agent plumbing to the UI people actually use.
Hebrew RTL mortgage-planning advisor
Full-stack, right-to-left Hebrew advisor that walks users through building a mortgage. A guided wizard models loan mixes and surfaces the tradeoffs, backed by a FastAPI service and Supabase Postgres.
Survival-chase browser game with an AI auto-player
A fast, cursor-controlled 2D survival game on HTML5 Canvas, you're a mouse dodging a clever cat across an arena, with Easy/Normal/Hard modes and an online Top-10 leaderboard. Every visual is drawn and every sound synthesized with the Web Audio API (no external assets). A companion userscript bot reads the live game state and plays to any target score, dodging each special attack.
Realtime Hebrew voice AI
Low-latency bridge between the browser and a live voice assistant: Ivrit speech-to-text → OpenAI Realtime → Cartesia text-to-speech, with LangGraph + Pydantic-AI guardrails sanitizing every reply and Langfuse tracing the whole conversation.
Pluribus-style 6-max poker AI
Autonomous No-Limit Hold'em bot for a bot-vs-bot competition. A Deep-CFR self-play blueprint feeds runtime lookup, exact real-time CFR river search, and a per-opponent exploitation layer, routing every decision to the layer that plays it best. 162 tests across 24 files.
AI tax platform, VLM extraction + RAG chatbot
The AI platform I lead the front end for (private / work). A VLM-powered reader extracts structured data from tax documents across multiple languages; an LLM support chatbot answers complex tax queries using RAG over vector search (LangChain + FAISS). Also delivered ~40% faster page loads.
Igal Babushkin · San Francisco Bay Area, CA
I'm an AI-focused full-stack engineer with a computer-science degree from Ben-Gurion University and three years shipping production software inside fast-paced startups. Recently relocated to the U.S. as a permanent resident, I'm looking for a team building real AI products.
My focus is applied LLM engineering. At SimpleTax I built a VLM-powered document reader that extracts structured data from tax forms across multiple languages, and an LLM support chatbot backed by RAG and vector search (LangChain + FAISS). I care about the whole path, retrieval quality, prompt and agent design, evals, and the interface that makes it usable.
Off the clock I keep shipping, realtime voice agents, game-theory bots, and plenty of automation. I move fast, wear many hats, and learn whatever the problem needs. If it ships and people use it, I'm in.
Contact
I'm open to full-time roles and interesting collaborations. The fastest way to reach me is email, I answer quickly.