Generative Al Engineer
Tata Consultancy Services (TCS) · Jan 2022 – Jan 2024
- Built a production GenAl microservice for semantic search over a legal knowledge base using vector similarity search through an internal Al search platform.
- Supported enterprise security hardening and formal internal security and quality certification requirements for the Al service.
- Containerized and deployed services across multiple environments, enabling consistent delivery on enterprise infrastructure.
- Built a chat-based development and test harness for orchestrating LLM agents and tool integrations across multiple LLM providers, accelerating agent experimentation and validation.
- Built an LLM-as-a-judge evaluation framework to benchmark RAG search quality and measure improvements across retrieval iterations.
- Built a centralized prompt management service for creating, managing, and versioning generative-Al prompts across multiple client applications.
- Built an LLM-driven document processing pipeline that extracts structured data from legal documents and converts it into datasets for evaluation and quality analysis.
- Built a multimodal mobile test automation agent using screenshots, Ul state, and vision-language reasoning to generate structured actions for mobile execution; refactored it into a tool-based architecture for extensibility.
- Researched Al-driven GUI grounding using OCR and fuzzy matching to locate Ul elements from natural-language instructions, including exploration of vision-model-based grounding approaches.
- Built ReAct-based, tool-calling agents with LangChain and LangGraph against an internal LLM gateway for incident-resolution and ticket-automation use cases.

