- 🤖 Freelance RAG / AI-Agent Engineer — I build production LLM systems for early-stage B2B SaaS.
- 🔭 Shipped: a Bedrock + FAISS RAG pipeline, a custom ReAct multi-agent system (no LangChain), and a published VS Code extension.
- 📊 I ship evaluated systems — RAGAS-scored, deployed, measured. Not demos.
- ⚙️ Core stack: Python · FastAPI · AWS Bedrock (Titan) · FAISS · RAGAS · OpenAI.
- 📫 Reach me at akash@akashchauhan.dev
Languages
AI / ML
Backend & Infra
AI / ML
Backend & Infra
| Project | Stack | What it does |
|---|---|---|
| 🌿 GreenLedger-AI | Python · AWS Bedrock · FAISS · RAGAS | Production RAG pipeline over domain docs. RAGAS: precision 1.0, recall 1.0, faithfulness 0.86 — evaluated, not assumed. |
| 🎯 Placement Pilot | Python · Custom ReAct · FAISS | Multi-agent placement-prep system. Hand-rolled ReAct loop (no LangChain), per-user FAISS retrieval. |
| 📚 Flashen | Python · LLM · SM-2 | Turns any PDF into flashcards with SM-2 spaced-repetition scheduling. |
| 🧩 Ayu TokPress | TypeScript · VS Code API | Published VS Code extension — 5-provider AI chat panel, live on the Marketplace. |
$ python ragas_eval.pycontext_precision (cases with >=1 relevant chunk) : 1.00 context_recall (answer found in top-5) : 1.00 avg token reduction : 9.7% faithfulness (LLM-graded) : 0.86
