Work Experience
Luminous Computing
May, 2019 - May, 2023
I was employee #2 and the Vice President of photonics at Luminous Computing, where I recruited and led a team of 9 engineers. Our team delivered the first monolithically integrated 112 Gbps PAM4 transceiver chips, fabricated in the 300mm GlobalFoundries Fotonix platform. I helped build the design software infrastructure (simulation, layout, integration) and the lab (including a home-grown 300mm wafer tester) from scratch. I also owned the key relationships with our photonics foundry partners (GlobalFoundries, SilTerra) and equipment vendors. The photonics IP lives on today at Enosemi.
Apple
June, 2017 - May, 2019
I helped reduce multi-radio coexistence interference in the Apple Watch Series 3 and 4 using both hardware and software techniques. I owned the automation software for in-factory testing of coexistence interference for these products.
PhD
2011 - 2017
I received my PhD from the Lightwave Lab at Princeton University, with a research focus on building photonic integrated circuits for ultra wideband wireless signal processing. For my thesis, I designed the first integrated circuit that reduced wireless interference over 1000x over every LTE channel in existence at the time, using analog interference cancellation. I co-founded a startup company called Rebeless to commercialize the technology. The company was not successful, and it shut down in 2017. At Princeton, I was a recipient of the National Defense Science and Engineering Graduate Fellowship and the Gordon Wu Fellowship.
Projects
2024 NFL Big Data Bowl Winner: Missed Tackle Opportunities
The Big Data Bowl is the premiere sports analytics data science competition hosted by AWS and the NFL. My team was selected as the grand prize winner from a field of over 300 teams using our metric, Missed Tackle Opportunities. Missed tackle opportunities represent a new class of defensive mistake that is not captured by the current statistics: think of defenders getting juked out, making ambiguous arm tackles, or being lazy. To detect these missed tackle opportunities, we trained a custom XGBoost model that ingests player tracking data to predict, in real-time, each defender's probability of making the tackle within the next second. We verified the model's accuracy by successfully identifying ~90% of the 1100x labeled missed tackles, and on top of that, also identified an additional 3500x missed tackle opportunities that were previously undetected. Missed tackle opportunities will be incorporated in the NFL's NextGenStats pipeline.
Writing
Pieces from my personal blog