Reimagining the radio by paying rideshare drivers to advertise music.So far:-built MVP, raised $400K pre-seed, built v1 product, launched on App Store, shipped to drivers :)Currently:-onboarding artists, scaling drivers, building artist business pipelines for scale

Developed CLI to extend Warden AIOps profiling data from agents running on 10,000's of Java applications including Salesforce CoreAPP.Implemented JavaFlightRecorder stack-frame caching for multi-pod and multi-tenant 24 hour+ time period aggregation of JFR log analysis from Warden agents.Warden Users can now fetch Warden data to local environments, setup regression identification on releases from warden logs, and automated nightly reports based on the profiling data.

I developed a vision-framework to filter out 98.6% of static oil field camera streams. Experimented with deploying Yolo V4 vs Faster R- CNN deep layers to achieve 92.6% precision detecting and preventing unauthorized activity and oilfield equipment failure.I also implemented RNN architecture with LSTM modules on Coriolis oil meter time series to detect gas blowby and breakout anomalies within EOG division oil separators (valued to prevent $3MM in operational losses).

At Ambrosia AI, we develop intelligent, cost effective medical devices to analyze and restore motion in the physically disabled—ranging from muscle soreness to post-stroke hemiplegia. Our flagship solution is a novel, portable, non-invasive, and cost-effective deep electroencephalography (EEG) based muscle stimulation frameworkConducted a 6-week clinical trial in India with cerebral palsy, post-stroke, and hemiplegic patients (180%+ average muscle improvement) while garnering $550K+ in receivable revenue from physiotherapy clinics/hospitals pending FDA approval.Our novel research has been awarded as a 2019 100 Global Google Science Fair Finalist and the 2017 Intel International Science and Engineering Fair (ISEF) finalist. We have also been received interviews from YCombinator and media outlets including CNN.

Worked on reinforcement learning research (building VRetas) at the Harvard Data Systems Laboratory with supervision from Stratos Idreos and Abdul Wasay.VRetas: Implemented a reinforcement learning system using Q-Learning and SARSA to learn how to efficiently visualize large data sets in virtual reality while accommodating for user preferences.Designed Action Set Reduction using Hierarchical Reinforcement Learning: Researched avenues towards a the development of a data system that uses data-fingerprinting and Hierarchical Reinforcement Learning to reduce and optimize the pool of possible actions.