Headline

Applied Generative AI Engineer @ Deloitte | MS AI & Cybersecurity @ IIT PATNA | Agentic AI Engineer | MCP · LangGraph | RAGAS + LLM-as-Judge | LLMOps: LangSmith + Langfuse + MLflow | Open to Global Opportunities

About

I’m an Applied GenAI Engineer with around 5 years of experience building enterprise GenAI systems across RAG, LLM-to-SQL, agentic workflows, Azure OpenAI, Databricks, LangGraph-style orchestration, FastAPI, MLflow tracing, evaluation, and LLMOps. At Deloitte, my recent work has focused on production-oriented GenAI platforms. I owned two major agents end to end inside Procurement GPT, a multi-agent procurement intelligence platform for a global FMCG client: • OKR Intelligence Agent — natural-language KPI/OKR analytics using controlled LLM-to-SQL over Databricks, with schema grounding, KPI metadata, business-rule resolution, sqlglot validation, semantic SQL validation, result-grounded responses, MLflow tracing, and UAT evaluation. • Knowledge Agent — enterprise RAG over internal procurement policies, SOPs, playbooks, supplier onboarding documents, compliance documents, and knowledge repositories, with ingestion, metadata, chunking, embeddings, Azure AI Search hybrid retrieval, semantic reranking, access-aware retrieval, citation validation, and no-answer fallback. I also led AI design for Studio, an AI avatar, voice, and script-to-video platform. The design followed a practical principle: LLM proposes, deterministic code disposes. The LLM handled scene planning and consent verification, while deterministic services controlled safety, cost, provider routing, credit reservation, consistency locking, retries, tracing, media assembly, and evaluation. My strength is building GenAI systems — with grounding, validation, access control, prompt-injection protection, observability, evaluation, cost control, and production-readiness. Core areas: Generative AI, Agentic AI, RAG, LLM-to-SQL, LangGraph, LangChain, Azure OpenAI, Databricks, Azure AI Search, FastAPI, MLflow, LLMOps, AI Agents, Vector Search, Semantic Reranking, Prompt Engineering, Guardrails, Python, SQL. I'm pursuing an MS in AI and Cybersecurity at IIT Patna alongside full-time work. The problems I find genuinely interesting sit at the intersection of agent reliability, evaluation, and observability: figuring out not just whether an agent worked, but *why* it worked, and whether it'll hold under adversarial conditions. Finally, I build production agentic AI systems, the kind that run inside enterprises, handle real data, and need to be right. 📧 ag.chaudhari.1512@gmail.com 🔗 github.com/Akash-1512 🌐 ai-akash.netlify.app

Location

Mumbai

AI stack

antigravity

codex

claudecode

Work experience

Data Scientist - GenAI / Agentic AI Engineer at Deloitte, Aug 2025 - Present · 1 yr 1 mo, Mumbai, Working on enterprise Generative AI and Agentic AI platforms using Azure OpenAI, LangGraph-style workflows, Databricks, Azure AI Search, FastAPI, MLflow, Azure App Service, and LLMOps practices. Key work: • Owned OKR Intelligence Agent and Knowledge Agent end to end inside Procurement GPT, a multi-agent procurement intelligence platform for a global FMCG client. • Built OKR Intelligence Agent for natural-language KPI/OKR analytics using controlled LLM-to-SQL over Databricks. • Designed the OKR pipeline across memory rewrite, intent/entity extraction, business-rule resolution, KPI semantic mapping, schema grounding, SQL generation, sqlglot validation, semantic SQL validation, Databricks execution, result validation, Excel export, and grounded answer generation. • Implemented governance for structured analytics using curated SQL views, KPI metadata, read-only execution, row-level authorization at the governed data layer, no-data fallback, prompt-injection testing, and MLflow tracing. • OKR Agent achieved around 93% on a curated 117-prompt client UAT benchmark. • Built Knowledge Agent for enterprise RAG over internal procurement policies, SOPs, playbooks, supplier onboarding documents, compliance documents, and knowledge repositories. • Designed the Knowledge Agent pipeline across ingestion, chunking, metadata, embeddings, Azure AI Search hybrid retrieval, semantic reranking, access-aware retrieval, citation validation, and no-answer fallback. • Led AI design for Studio, an AI avatar, voice, and script-to-video platform in staging and production-readiness testing. • Designed Studio components including LLM scene planner, spoken-consent verification, locked identity/voice/wardrobe/background references, engine registry, credit calculator, reserve-before-spend flow, async worker pipeline, SSE progress updates, and final 9:16 MP4 assembly.

PC GenAI - Analyst at Government of Maharashtra (GoM), Dec 2021 - Jun 2025 · 3 yrs 7 mos, Maharashtra, India, Worked on NLP, retrieval, and Generative AI workflows for public-sector knowledge systems and citizen-facing applications. Key work included: • Developed NLP and Generative AI pipelines covering text preprocessing, chunking, retrieval, prompt engineering, and evaluation. • Built embedding-based retrieval systems using FAISS and Chroma to ground LLM responses in authoritative government documentation. • Applied hybrid retrieval strategies to improve response grounding and reduce hallucination risk in compliance-sensitive use cases. • Designed document ingestion and indexing pipelines for continuous knowledge base updates aligned with policy and regulatory changes. • Used RAGAS-style evaluation across faithfulness, context relevance, and groundedness to improve retrieval and generation quality. • Collaborated with policy and technical stakeholders to convert domain requirements into deployable AI solutions with traceable outputs.

Education

Master of Science - MS, Artificial Intelligence and Cyber Security, Indian Institute of Technology, Patna

Bachelor of Engineering - BE, Automobile engineering, Mumbai university