Dwijen Chawra
I'm a junior studying Computer Science at Purdue University. My interests lie in the intersection of ML and high performance computing.
Previously, I interned at Arta Finance, where I optimized compute resource usage and load times for data science workflows, engineered an automated financial advisor, and automated local volatility models for exotic options pricing.
I was an undergraduate researcher at the Zeitzer Circadian Research Lab at Stanford University, where I worked on developing models to predict sleep quality from polysomnography data.
What I'm up to
- Undergraduate teaching assistant for CS 252: Systems Programming.
- Training bimanual robots to learn from demonstration at Purdue MARS Lab.
Projects
Wave
GitHub | Aug 2022 - May 2023
Gesture interactions with a computer through user-trained hand gestures with only a webcam. Sub-second latency for action recognition, and scales to infinite gestures.
Drug Identification with NER Models
Project Page | GitHub | Aug 2022 - May 2023
Worked with Battelle Inc. to develop a NER model to identify drug name and usage from medical records, utilizing a BioBERT bi-directional LSTM model to achieve 94% accuracy on the test set.
Soundscape
GitHub | Jan 2023
Created an intelligent playlist generator that uses NLP and performs sentiment analysis on transcribed speech. Stored vectorized songs in CockroachDB's relational database and generated playlists based on user emotion.
Publications
A threshold by any other name: is five minutes of wake 'long' enough to degrade sleep quality?
Renske Lok, Dwijen Chawra, Jamie M Zeitzer
The Impact of Missing Data and Imputation Methods on the Analysis of 24-Hour Activity Patterns
Renske Lok, Lara Weed, Dwijen Chawra, Joe Winer, Jamie Zeitzer