Boosted accuracy of ICD-10-CM medical coding from 54% to over 90%.

Implemented a reasoning layer for the medical coding workforce.

Boosted customer trust, leading to a deal worth $2M in ARR.

Designed an application using Google’s CodeGemma LLM and LLVM, reducing time to generate fuzz targets by 5x.

Hardened Pure's NFS v4.1 file system by uncovering and fixing 7 high-severity security vulnerabilities.

Led a team of three engineers and won company's Global GenAI Hackathon.

Created a transformer-based ML model for WBC hunting, improving recall by 12%.

Led ML development lifecycles for WBC cell hunting and monolayer detection.

Developed a scanning algorithm for whole-slide imaging, reducing scan time by half.

Designed database schema and endpoints for new features including Cases and Workflows using Sequelize.

Contributed to back-end Node.js/Express software, securing RESTful APIs and developing new features like Cases.

Delivered critical ML lifecycles for WBC cell hunting, monolayer detection, and out-of-focus image detection.

Mentored junior engineers in designing RESTful APIs and set up ML training, deployment, and monitoring infrastructure.

Developed a low-latency video streaming infrastructure by building on top of the WebRTC native library in C++ and hosting TURN servers.

Selected and fine-tuned an ML model for person fall detection, improving accuracy compared to production by 5%.

Built a custom WebRTC real-time video streaming web platform deployed in over 5 Fortune 100 companies.

Scaled a Kafka event processor by 40% (to 1 billion daily events or 11,500 QPS) by restructuring systems and load balancing.

Optimized distributed systems' performance by 25% through batch processing of SQL queries and horizontal scaling.