I'm Praveen Devarajan, a Computer Science/Public Policy grad student at the University of Chicago ๐
-
I work as a Data Scientist/Research Engineer with UChicago Institute for Climate and Sustainable Growth, where I help develop ML products to detect violations of the Clean Air Act. I am also a DRW fellow, through which I assist in understanding the effect of Time-Of-Use pricing of electricity on consumer behavior.
-
I worked as a Data Engineer with Ballotpedia, where I built a pipeline to extract, clean, and load data for more than 100,000 candidates contesting local elections in the US.
-
I also help build the backend of a sustainability-based startup called Solom.
I have been a volunteer teacher, a grassroots social worker during the pandemic, and a product manager helping build products for teaching children.
- Web scraping and Tokenizing pipeline built using Go
- Pipeline to extract, parse and classify Twitter users
- Energy sector analysis of US
- Belay, a slack-like messaging app
- AI-based Job Application tracker(ongoing)
Many repositories from school projects are private for academic integrity and intellectual property purposes. Code can be shared with prospective employers upon request.
- Distributed System for Genomic Annotation Services using AWS Services(S3, Glacier, DynamoDB, State Machines, Lambda functions and SQS/SNS)
- Lock-free task management system in Go, utilizing concurrent data structures (lock-free FIFO queue, RWLock) and atomic CAS operations
- Parallelizing image convolutions using basic synchronization primitives
- Text to text Transfer Transformer model (T5), fine-tuned to summarize data
- Replicating the "Attention is all you need" architecture using PyTorch
- Markov Models for Text Attribution
- Web Scraping & Web Crawlers with lxml
- CRUD app to handle tweets, and link records, with SQLite and Flask ๐ชถ
- Record Linkage Algorithms
- Analyzing Avian Diversity using Recursive Treemaps
- Polling Precinct Simulation using M/M/N Queues
- Analyzing Political Tweets using NLP
- Epidemic Simulation using the SIR Model