Headline

AI Engineer @AuxoAI | IIT Delhi | JSW

About

I’m an AI Engineer focused on building reliable, production-ready AI systems across agentic AI, LLM applications, RAG, knowledge graphs, and backend infrastructure. At AuxoAI, I build production AI systems using LangGraph, Temporal, PostgreSQL, Neo4j, and LLM observability tooling. I’ve worked on durable agent workflows, 10M+ row data systems, intent-routed knowledge graph tools, ontology extraction pipelines, source-traceable RAG, and end-to-end LLM tracing. Some of my recent work includes reducing redundant agent tool calls by 30% and average query latency from 90s to 52s, reducing failure isolation time from 2 hours to 15 minutes, and onboarding 15+ domain ontologies with zero code changes. I also build AI-focused projects involving email automation and multi-speaker audio scoring using diarization, ASR, GPT-4o, Whisper, and Gemini. I enjoy working at the intersection of AI and engineering—especially making AI systems reliable, observable, scalable, and production-ready. My core stack includes Python, SQL, TypeScript, LangGraph, OpenAI, Gemini, FastAPI, GraphQL, Temporal, PostgreSQL, Neo4j, AWS, GCP, and OpenTelemetry.

Location

Bengaluru

AI stack

anthropic apis

langfuse

langgraph

openai

Arize Phoenix

LiteLLM

Work experience

AI Engineer at AuxoAI, Jul 2025 - Present · 1 yr 3 mos, Built agent-based lineage and migration workflows using LangChain on DeerFlow with Langfuse for tracing and evaluation; parsed 30+ Oracle ODI SQL scripts (1500+ lines each) into column-level lineage and automated migration of 150+ Greenplum procedures (2000+ lines each) to Teradata, improving accuracy from 75% to 92% and reducing manual effort. Built custom ontology-based knowledge graph pipelines with citation-aware RAG using Unstructured.io and Docling; governed rendering, domain grouping, and edge semantics through a single-source ontology YAML decoupled from transform logic, with consistent source tracking and provenance across vector and graph DBs. Architected an end-to-end data lineage visualization platform — Neo4j graph backend, FastAPI gateway, and D3/Dagre frontend — with hybrid retrieval combining vector search on node embeddings and BFS Cypher expansion, supporting 4 upstream graph producers behind a unified rendering pipeline. Designed an AgentMiddleware layer intercepting all tool call results to automatically surface output file paths as thread artifacts — including self-contained Chart.js HTML companions for JSON outputs — retroactively improving every existing and future agent without requiring changes to agent code.

Software Engineer Intern at JSW Group, May 2024 - Jul 2024 · 3 mos, Built an HR chatbot using RAG with Llama3 and LangChain, implementing multi-source data ingestion and vector embeddings to improve context retrieval and response accuracy. Deployed and integrated the solution with Darwinbox in collaboration with HR & IT, reducing workload by 50 man-hours/day; received Letter of Recommendation from Head of IT for contribution.

Summer intern in IT & Digital Solutions Team at JSW, May 2024 - Jul 2024 · 3 mos, Mumbai, Maharashtra, India, - Worked with 25+ HR professionals to design and deploy an interactive chatbot, improving query resolution and customization - Incorporated Meta Llama 3.1 with 8Bn+ parameters, fine-tuned using RAG method, and optimized model's performance - Achieved 87% accuracy, cutting 35+ man-hours per month of HR; Resulted in a group-wide integration of the model Received Letter of Recommendation from Head of IT & Digital Solutions, JSW Mumbai for exemplary work accomplishment

Coordinator at Rendezvous IITD, Aug 2023 - Oct 2023 · 3 mos, New Delhi, Delhi, India

Analytics Intern at Gayatri, May 2023 - Jul 2023 · 3 mos, Uttar Pradesh, India, Analyzed & visualized cleaned data to deliver insights, enhancing decision-making at Gayatri Projects' head office in Hyderabad.

Education

Bachelor of Technology, Mechanical Engineering, Indian Institute of Technology, Delhi

Public links

github: https://github.com/BruceWayne4