Headline

AI Engineer | Agentic AI & Multi-Agent Systems | RAG, LLMs, LangGraph | Python, FastAPI | Software Engineer @ Wipro

About

I build AI systems that survive production, not just demos. For the past year at Wipro I've shipped agentic and GenAI systems on QASuite, an enterprise QA platform — working on the unglamorous parts that decide whether an LLM system actually holds up: agent orchestration, retrieval quality, and running models on infrastructure you control. What I've shipped: → Re-architected the core workflow from a linear pipeline into a LangGraph cyclic state machine with human-in-the-loop approval gates. 40% better scalability, and 70% of test-case generation automated through LLM-based semantic analysis → Stood up self-hosted LLM infrastructure (Ollama serving DeepSeek and Gemma), owned end to end from model selection to offline serving, kept 100% of data on-prem and removed cloud inference cost entirely → Built an autonomous, webhook-triggered agent wired into Jira, GitHub and Azure DevOps, cutting end-to-end test evaluation and validation time by 50% → Currently leading the decomposition of the monolith into microservices, with agents and core services as independently deployable units Built outside work, same class of problem: CI Researcher : a multi-agent competitive intelligence system. LangGraph with Send() for parallel, isolated per-company research, human-in-the-loop approval before report finalization, a credibility-scoring engine, and a FastAPI backend streaming live agent progress over SSE into a Next.js 14 dashboard. The hardest part wasn't the agents: it was a RAG failure where ChromaDB silently returned 0 chunks from a client/server ID mismatch, which I fixed by redesigning ingestion around server-generated session IDs with automatic cleanup. StockBot : a RAG chatbot for stock-market reasoning, built on FastAPI with FAISS and AstraDB for vector search. My edge: I came to AI from software engineering, so I treat agents as systems to be debugged, deployed and measured, not prompts to be tuned. Currently open to AI Engineer, GenAI Engineer and Agentic AI Engineer opportunities in Bengaluru or remotely within India. Ideally on a team building AI-native products rather than bolting AI onto existing ones. Open to connecting with anyone working on agents, RAG and applied AI in production. anuragg0582@gmail.com

Location

Bengaluru

AI stack

langgraph

mcp

githubcopilot

Chroma

Pinecone

Tavily

Work experience

Software Engineer at Wipro, Jun 2025 - Present · 1 yr 6 mos, Bengaluru, Karnataka, India, Building production GenAI and agentic systems on QASuite, Wipro's AI-driven QA platform. • Re-architected the core workflow from a linear pipeline into a LangGraph cyclic state machine with human-in-the-loop approval gates — improved scalability by 40% and automated 70% of test-case generation via LLM-based semantic analysis. • Leading the modularization of a monolithic codebase into a microservices architecture, decoupling agents and core services into independently deployable units; partially shipped to production. • Deployed and operated self-hosted LLM infrastructure (Ollama serving DeepSeek and Gemma), guaranteeing 100% data privacy and eliminating cloud inference costs — owned the full lifecycle from model selection to offline serving. • Built an autonomous, webhook-triggered AI agent integrated with Jira, GitHub and Azure DevOps, cutting end-to-end test evaluation and validation time by 50%. LangGraph · LangChain · Python · FastAPI · Ollama · RAG · microservices · Docker

Trainee at Wipro, Apr 2025 - Jun 2025 · 3 mos, Bengaluru, Karnataka, India

Software Engineer Intern at Pet Station Technologies, Jan 2024 - Jun 2024 · 6 mos, India

Education

Bachelor's degree, Electrical, Electronics and Communications Engineering, Indian Institute of Information Technology, Design and Manufacturing, Jabalpur

Public links

github: https://github.com/shady528

leetcode: https://leetcode.com/shady528