GenAI Engineer
I build multi-agent systems, RAG pipelines, and ship open-source tools that make AI development faster. Currently at Quantiphi.
I work at the intersection of large language models and production engineering. At Quantiphi, I've built multi-agent systems that cut pipeline runtimes by 95%, shipped NL2SQL agents for enterprise platforms, and optimized inference costs by 80% through prompt restructuring and caching strategies.
Outside work, I ship open source — my tool patchwork-conventions is published on PyPI and MCP Registry, and integrates with Claude Code, Cursor, and GitHub Copilot.
1.5+
Years Experience
95%
Runtime Reduction
4
Certifications
Quantiphi · Bengaluru, India
Jul 2025 - Present
Developed in 2 days. Published to PyPI and MCP Registry. Integrates with Claude Code, Cursor, and GitHub Copilot.
AST-based codebase scanner that auto-generates CONVENTIONS.md for AI coding agents. Tree-sitter analysis detecting naming conventions, import patterns, error handling, testing frameworks, and API shapes across 5 languages with confidence scores and real code examples.
95% runtime reduction — 45 min down to 5 min via dynamic parallel orchestration.
Multi-agent staffing system using Google ADK that takes a project ID or role description, dynamically spawns parallel agents for each required role, and returns top candidate recommendations with fit rationale.
Enables non-technical users to query complex datasets using plain English.
Conversational NL2SQL agent that lets PMs and business leads query project health data in plain English, converts it to BigQuery SQL via an MCP server, and returns summarized results.
Used by 2,000+ employees. Quantifies developer productivity gains from AI-assisted coding.
Developer tool used by 2,000+ employees that captures AI interaction logs from Copilot, Kiro, and Claude Code, scores prompt quality using LLM analysis, and calculates effort-saved metrics per developer.
Internal developer tool used across engineering teams at Quantiphi.
Audited and improved the internal AI coding assistant (CLI + VS Code extension) by benchmarking against open-source tools, rewriting system prompts, and adding slash-command based skill invocation.
Processes 2,000+ project documents. Powers the knowledge base used by production AI agents.
Ingestion pipeline that pulls 2,000+ project documents from Google Drive, summarizes them via LLM, and indexes into Vertex AI Vector Search for the staffing agent's RAG retrieval.
Cut daily inference cost from $4,000 to $2,500 (37% reduction).
Ran implicit vs. explicit caching POC and restructured all agent system prompts. Set up dynamic LLM routing across Gemini and Claude model families to pick cheaper/faster models based on task complexity.
SRM Institute of Science and Technology
B.Tech in Computer Science Engineering
I'm always open to discussing AI engineering, multi-agent systems, or interesting collaboration opportunities. Feel free to reach out.
Say Hello