Delta Kernel Rust (2025 – Present)

Core contributor to the open-source Rust implementation of the Delta Lake storage engine.

• Designed and implemented core read/write optimizations, reducing query latency by up to 2.5x and unblocking multiple downstream engine integrations

• Overhauled table feature representation and protocol validation, improving extensibility and correctness

• Investigated and resolved subtle correctness bugs spanning serialization, memory layout, and cross-engine compatibility

Managed Storage (2023-2025)

Built cloud storage infrastructure powering Databricks' multi-tenant platform across AWS, Azure, and GCP.

• Delivered scalability improvements enabling 100k+ customers with sub-200ms latency and 100% uptime at peak

• Redesigned core reconciliation system, reducing database transactions by 55x and lock acquisition time by 35x

• Led full-scale load testing at 150k+ buckets, identifying and resolving database bottlenecks that restored P95 latency from 2,000ms to 400ms

• Rearchitected bucket provisioning and lifecycle management, achieving ~10x latency improvement and eliminating multi-hour rollout delays

• Master's research in Berkeley Speech Group, focusing in Differentiable Digital Signal Processing

• Developed a novel real-time data efficient machine learning model for articulatory synthesis

• Engineered a streamlined architecture requiring only 0.4M parameters (vs. 9M baseline) while improving Word Error Rate by 1.63% (to 6.67%) and Mean Opinion Score by 0.16 (to 3.74)

• Achieved 4.9x inference speedup on CPU compared to State Of The Art models while maintaining speech quality, enabling real-time applications

• Designed and implemented a critical test infrastructure for satellite hardware torque and energy margin verification, mitigating flight risks and reducing test time by 10x.

• Developed a spectral analysis system leveraging Short-Time Fourier Transform (STFT) to detect and characterize reaction wheel bearing faults.

• Created a comprehensive test automation suite for flight component qualification, integrating hardware control interfaces and data acquisition systems.

• Architected and implemented full-stack tooling to automate ML product private offer creation in AWS Marketplace, reducing manual engineering effort from hours to minutes

• Built RESTful APIs and implemented database integration for offer management system

• Established robust CI/CD pipeline with automated integration testing, significantly improving development tool reliability

• Led the design and construction of a custom 4-antenna array system for tracking Lincoln Experimental Satellites (LES), achieving successful signal capture at 230 MHz from multiple satellites

• Developed and optimized a Blind Adaptive Beamforming algorithm to isolate and identify satellite signals amid complex co-channel interference, improving signal-to-noise ratio by 30%.

• Implemented end-to-end integration testing by using phase shifters and RF components to simulate antenna array geometry, and capturing and processing data using GNURadio, Python, and MATLAB.