MSc Data Science student at SRH University Heidelberg. I build production-grade AI systems, not just notebooks.

SAS2Py - AI-powered SAS-to-Python translator built as a POC for Deutsche Bank. Fine-tuned 3 LLMs (Qwen 1.5B, DeepSeek 6.7B, Llama 8B) with LoRA and QLoRA. Hybrid RAG pipeline with pgvector. 100% syntax validity on the smallest model. FastAPI + Streamlit + Docker + 75 automated tests.

Ethical Credit Scoring System - GDPR/EU AI Act/German AGG-compliant credit scoring with Random Forest, 3-stage fairness mitigation, SHAP explainability, Llama 3.3 70B AI briefings, and a live Gradio dashboard for credit officers.

Multi-Agent RAG Pipeline - 15-agent system using LangGraph for cross-document Q&A over OECD/IMF/UN policy PDFs. Hybrid retrieval (BM25 + Pinecone), debate agents, NLI-based fact checking.

Python, PyTorch, Hugging Face, LangChain, LangGraph, LoRA/QLoRA, RAG, FastAPI, PostgreSQL, pgvector, Docker, SHAP, scikit-learn, spaCy

• Looking for an internship or Werkstudent in GenAI, ML, NLP, or Data Science in Germany
• Based in Heidelberg, available immediately
• sarveshmmokal@gmail.com | LinkedIn