Rahul Rachhoya

AI Engineer with 3+ years building production LLM systems (Claude, GPT-4, Mistral) serving 10K+ users. Specialized in RAG, voice AI, multi-agent workflows on AWS. 40% cost reduction, 35% hallucination improvement, 92% model accuracy.

AI Engineer - Voice AI & Conversational Systems - Careers360 - Hyderabad, India

(2026-01)

• Architected voice AI career counselor using Claude/Bedrock + Sarvam AI (STT/TTS) + pgvector serving 10K+ sessions; RAG pipeline reduces hallucinations 35%
• Optimized costs 40% ($0.40→$0.24/session) via prompt caching + A/B testing; LangSmith observability tracking latency, cost, quality
• Built multi-agent workflow (function calling) improving accuracy 28%; AWS Lambda deployment with 99.7% uptime

AI Engineer - Crystaltech Services - Remote, India

(2024-08 - 2025-12)

• Built RAG pipelines reducing hallucinations 35%; hybrid search (BM25+vector) + Cohere reranker improving retrieval 85%→91%
• Fine-tuned Mistral 7B using LoRA achieving 92% accuracy (vs 78% base) with 40% faster inference; training cost $50 vs $2K
• Weights & Biases monitoring 50+ LLM iterations; A/B tested GPT-4 vs Claude achieving 60% cost savings
• Built 4-agent document system (LangGraph) processing 500+ docs with 94% accuracy; 75% time reduction (30min→7min)

System Engineer (AI/ML Focus) - STL Digital Limited - Pune, India

(2022-06 - 2024-07)

• Implemented RAG system improving accuracy 40% (65%→91%) using ChromaDB + GPT-3.5; 500+ daily queries with 1.8s response
• Built AI-assisted validation reducing errors 60%; deployed Streamlit dashboard to AWS serving 500+ daily users
• Implemented pytest suite with 85% coverage; CI/CD pipeline via GitHub Actions

Master of Computer Applications (MCA) - AI, Data Science, Machine Learning - University of Hyderabad (2019 - 2022)