About Me

Georgia Tech

BITS Pilani, Goa Campus

DoorDash

• Developed an LLM-based agentic copilot for Dasher earnings optimization and real-time support across on-demand resolution workflows in latency-sensitive logistics environments.
• Architected a self-healing pipeline combining multi-axis LLM-as-judge evaluation, trend anomaly detection, and structured root-cause analysis. Routed prompt-level fixes through GEPA optimization and code-level issues to human triage; the pipeline was actively used during copilot development.

Microsoft Research

• Developed and evaluated novel algorithmic solutions for large-scale code intelligence tasks, applying advanced engineering principles to improve code QnA, summarization, understanding, and translation across gigabyte-scale Microsoft repositories.
• Engineered state-of-the-art LLM algorithms featuring parametric/non-parametric memory, reinforcement learning fine-tuning, and advanced RAG-based strategies (GraphRAG, HippoRAG, PlanRAG), achieving 5% gains in code translation accuracy.

Microsoft Research

• Designed and implemented an agentic + symbolic AI system that generates engineering diagrams directly from natural language, integrating principles of object-oriented programming and advanced AI/ML techniques.
• Collaborated with senior researchers, executives, and stakeholders at Microsoft Energy to refine feature requirements and ensure AI solutions met industry standards and responsible AI practices.
• Presented findings at the AAAI 2025 Workshop on AI to Accelerate Science and Engineering, demonstrating robust time management and the ability to quickly assimilate emerging AI methodologies.