cv

Technical Head

STEALTH STARTUP

End-to-End design and development of a dev tool for monitoring, analysing and optimizing AI agent

• Led the development of a product for real-time analysis and optimization of multi-agent AI systems
• Implemented KPI-driven prompt optimization, improving task success rate by 25% through iterative tuning and evaluation
• Integrated support for popular frameworks including LangChain, CrewAI, and AutoGen
• Used by 5 teams across multiple early-stage startups for debugging and optimizing AI agents

Senior Member Technical Staff

Oracle India Private Limited

End-to-End design and development of a cloud native service for data consistency across microservices

• Orchestrated distributed data consistency by developing and launching a cloud-native Go service – MicroTx
• Scaled the system to process over 300 billion transactions at 8,500 transactions/sec, with failure rate below 0.0001%
• Led a team of three developers
• Engineered strong data consistency via industry-standard XA protocol supporting MySQL, Postgres, and Oracle DB
• Implemented eventual consistency through LRA and TCC protocols for stateless NoSQL DB
• Built unified Helm charts to enable seamless deployment on AWS, Oracle Cloud, and GCP Kubernetes environment
• Developed admin console backend APIs to support real-time transaction monitoring, tracing, and service health checks
• Adapted Helm deployment patterns to support cloud blockchain provisioning in 10 minutes across cloud environments

Developer

SAP Labs India Private Limited

Implemented server-driven UI system and developed tools for Enterprise Application Development

• Led the design of Server-driven UI system that automated front-end generation and eliminated manual development
• Reduced delivery timelines from 12 weeks to 4 weeks while saving 40% resources of 75 engineers
• RESTful Application Model: Built tools adopted by 10,000+ engineers to accelerate ERP app development
• Auto-generated core entities and all service artifacts from API specifications to consume remote services
• Reduced API and UI validation cycle time by over 90%, bringing it down from an hour to five minutes
• Implemented fine-grained access control for Data Preview, enabling secure exploration of 10,000+ S/4 HANA tables
• Recognized with Best Early Talent Award in 2018 and 2019 for top performance and high-impact contributions

Master of Science in Computer Science

New York University, Courant Institute of Mathematical Sciences

3.8/4.0

• Honors Analysis of Algorithms
• Artificial Intelligence
• Deep Learning
• Computer Vision
• Reinforcement Learning
• Conceptual Gaps in Modern Machine Learning
• GPUs
• Big Data Application Development
• Distributed Systems
• Operating Systems
• Foundations of Entrepreneurship

Bachelor of Engineering in Computer Science and Engineering

PES University

3.6/4.0

• Data Structures with C
• Unix
• Operating Systems
• DBMS
• Software Architecture
• Computer Architecture
• Systems Software
• Software Engineering

Catalyst

SAP Labs

Awarded the SAP Catalyst recognition for outstanding contributions, exceptional learning ability, and potential to excel in critical roles

Catalyst

SAP Labs

Awarded the SAP Catalyst recognition second year in a row

Reinforcement Learning in Rocket League (RL²) Multi reward system

A deep dive into how reward design shapes the instincts of AI agents on the Rocket League field.

Dynamo DB - Effectiveness of Anti-Entropy and Gossip Protocol

A DynamoDB-inspired distributed key-value store that explores how gossip, anti-entropy, and quorum settings impact eventual consistency and fault recovery.

Image Geolocation

Developed an advanced image geolocation system using CLIP and RAG techniques

• Implemented novel hierarchical feature clustering approach
• Achieved up to 85.08% accuracy at continent level and 15.01% at street level

Cola Kernels

custom CUDA kernels for Compositional Linear Algebra framework}, focusing on Arnoldi iteration, Cholesky decomposition, Hutchinson trace estimation, and SVD

• Achieved up to 5x speedup and reduced memory usage by 8x compared to PyTorch implementation.

Sentiment And Capital, The Data Behind Financial Movements

Analyzes financial data and sentiment, revealing interconnections across stocks, Bitcoin, loans.

Future Video Frame Prediction

Predicted the 22nd frame of a video featuring moving objects using UNet architecture and Masked Conditional Video Diffusion

• Secured top-five position among 20 teams