AI Engineer · Agentic AI · Generative AI · AI-powered backend systems
Bengaluru, India
Explore my portfolio ↗ · LinkedIn · Email · Résumé
I see a repetitive task and start thinking: can we make this smarter?
I experiment with AI agents, build practical automation, and turn real-world problems into systems that help people get more done. I enjoy the work between an ambitious idea and something people can actually use.
Currently an AI Engineer at McLaren Strategic Solutions, building code intelligence, evidence-grounded documentation, and specification-driven modernization workflows. Previously a Senior Software Engineer at Tata Elxsi, working on AI-assisted testing, RAG, and cloud automation.
| Area | How I approach it |
|---|---|
| Agents & automation | Connect models to tools, relevant context, and validation so they can support real engineering tasks. |
| Knowledge & code intelligence | Use graphs, source retrieval, and lineage to make complex systems easier to understand. |
| Backend & delivery | Build with Python, APIs, and cloud automation to turn useful ideas into working systems. |
01 / Understand the system, then modernize it
At McLaren, my work connects application code, databases, reporting, and ETL into a Neo4j knowledge graph. The same foundation supports evidence-based documentation and component specification packs with dependency-aware implementation plans.
- Analyzed a legacy estate of approximately 30M lines of code, with a graph of approximately 2.4M nodes and 3.9M relationships.
- Combined MCP graph tools with targeted source retrieval to ground technical explanations.
- Used deterministic graph logic for component discovery and scheduling, with LLM-assisted behavioral descriptions and repair gates for incomplete output.
02 / Let agents take on repetitive testing work
At Tata Elxsi, built a CrewAI and Gemini unit-test generation framework for Python and C++, covering creation, validation, and coverage analysis. Reported approximately 60% less manual testing effort.
03 / Make knowledge accessible and setup repeatable
Built enterprise RAG workflows with Mistral 7B and customer-cloud deployment, alongside AWS EC2 orchestration that reduced manual setup effort by approximately 70–80%.
These are summaries of professional work. Enterprise source code is not public; scale and impact figures are approximate. Generated specification packs and test scaffolds still require implementation and behavioral validation.
Read the case studies → · View the portfolio source →
| Technologies & practices | |
|---|---|
| Core engineering | Python · C++ · FastAPI · REST APIs |
| AI & agents | LLMs · MCP · CrewAI · LangChain · Prompt engineering |
| Retrieval & graphs | RAG · Neo4j · Cypher · Hugging Face · Code & data lineage |
| Cloud & quality | AWS EC2 · AWS Bedrock · Docker · GitHub Actions · Pytest |
- Extra Mile Award — LLM-agent enhancements to unit-test automation at Tata Elxsi.
- Customer appreciation — secure RAG delivery in a customer cloud environment.
- Publication: Accelerating Unit Testing using Generative AI — Tata Elxsi, October 2024.
- Delivered internal GenAI training and enjoy sharing practical lessons from engineering.
Travel stories, guitar, gym, and swimming keep me exploring outside work. I capture moments while travelling and narrate them on HarshSaga.
I also enjoy tech events, meeting people building with AI, and hearing perspectives that challenge my own. There is always something new to learn.
Have a real problem worth solving?
Open to conversations about AI engineering, agentic AI, GenAI, and AI-powered backend roles.
Let's connect ↗