AI Engineer based in Berlin. I build production LLM agent systems — voice and chat agents, RAG pipelines, and the retrieval and orchestration plumbing that makes them reliable in the wild.
Currently building applied LLM systems in production. Off-hours I write about system design and turn research ideas — diffusion, clustering, rare-event learning — into small, readable repos.
- ai-phone-assistant — Production-grade healthcare phone agent. Google ADK, FastAPI, Kubernetes.
- Trip-Planner — Autonomous AI agent that plans end-to-end trips.
- feynman-technique — A Claude/Codex skill that turns resolved engineering work into reusable learning notes.
- system-design-playbook — Mental models, tradeoffs, and failure modes for system design.
- variational-fuzzy-kmeans — Soft clustering via a variational free-energy objective.
- entropy-afl-reranker — Entropy-regularized focal filtering and calibrated reranking for rare-event classification.
LLM agents (Google ADK, LangGraph) · RAG / pgvector · Python · FastAPI · TypeScript · PostgreSQL · Temporal · Docker · Kubernetes · GCP
- Medium — https://medium.com/@anotherBadCode
- LinkedIn — https://www.linkedin.com/in/abhishek-sharma-50254721/



