- 🎓 Final-year B.Tech CSE (AI & ML) student at B.V. Raju Institute of Technology (Expected May 2027).
- 🔬 Research Assistant specializing in Multilingual RAG architectures and benchmarking vector databases.
- 🌐 Former Campus Ambassador for Perplexity AI.
- 📄 Corresponding author of NutriRAG-ML, a multilingual RAG framework — accepted at CML 2026, Springer (Scopus-indexed).
- 🔭 Building ARGUS, an 11-agent hierarchical due-diligence system on Google's Agent Development Kit (ADK) for my thesis.
- 🔧 Merged a PR to Docling (IBM Research, 64k+ ⭐), fixing a hyperlink-extraction regression.
- 📫 Reach me at: rishabhsanu11@gmail.com
🕵️♂️ ARGUS | Multi-agent LLM system for automated financial due-diligence
- Architecture: 11-agent hierarchical system on Google's Agent Development Kit (ADK) with runtime orchestration and parallel tool-calling.
- Execution: Features a dedicated code-execution agent that computes financial metrics using real Python/pandas instead of relying on LLM estimation.
- Security & Eval: Includes a deterministic groundedness verifier, human-in-the-loop approval gates, 4 prompt-injection guardrails, and an 8-scenario evaluation harness (90 automated tests).
🥗 NutriBot (NutriRAG-ML) | Multilingual RAG for personalized diet recommendations
- Tech Stack: Python, FastAPI, ChromaDB, Hugging Face Transformers (all-mpnet-base-v2), Gemini 2.5 Flash, Cross-Encoder reranking.
- Impact: Grounded responses in a 74-item knowledge base across English, Hindi, and Telugu. Achieved a 1.6% hallucination rate with 0% safety violations.
- Recognition: Published at CML 2026, Springer (Scopus-indexed).
🛠️ Docling (Open-Source Contributor) | IBM Research
- Impact: Root-caused and fixed a hyperlink-extraction regression dropping URLs across ODT paragraphs, headings, and list items (PR #3949, merged).
- Quality: Shipped through 3 rounds of maintainer review with zero regressions across 44 passing tests.
Languages & Core:
AI / ML / Data Science:
Backend & Infrastructure:
