Looking for a job

Aditya Purohit

@adityapurohit3

AI/ML Systems Engineer at TCS Research & Innovation Labs who architected multi-agent workflows with a 99.4% task completion rate.

India

Message

At TCS Research & Innovation Labs, I architected multi-agent research workflows using LangGraph and Google ADK, coordinating specialized agents across 500K+ runs with a 99.4% task completion rate.

I engineered GraphRAG and hybrid retrieval pipelines over 500+ unstructured corporate records using Neo4j and PostgreSQL/pgvector. These pipelines improved precision by 35% and recall by 28%.

I also built evaluation harnesses with LangSmith and designed an event-driven research platform using GCP Cloud Run and Pub/Sub, cutting research turnaround by ~60%.

Earlier, as a Software Engineering Intern at Twelve 21, I built data ingestion and semantic search pipelines across 1M+ enterprise records, and developed asynchronous REST APIs and microservices. In my Document Copilot project, I built an investment research platform for SEC 10-K filings with hybrid retrieval and deep-linked audit citations.

Experience

Work history, roles, and key accomplishments

Software Engineering Intern

Twelve 21

Built automated data ingestion and semantic search pipelines across 1M+ operational enterprise records using Snowflake and SQL, streamlining incident classification, entity extraction, and automated normalization. Developed high-concurrency asynchronous REST APIs and microservices using Python (FastAPI) and PostgreSQL, handling relational schema modeling, connection pooling, and Redis caching.

Snowflake SQL Python fastAPI PostgreSQL Redis REST APIs Microservices GitHub Actions CI CD

AI/ML Systems Engineer

TCS Research & Innovation Labs

Architected multi-agent research workflows using LangGraph and Google ADK, coordinating specialized agents across 500K+ runs with a 99.4% task completion rate. Engineered enterprise GraphRAG and hybrid retrieval pipelines over 500+ unstructured corporate records using Neo4j and PostgreSQL/pgvector, driving a 35% gain in precision and 28% in recall.