Learning Structured Sparce Recovery
- Proposed, implemented Q-leaning agent to recover model for sparse compressive sensing.
- Agent trained for 15k iterations outperforms the individual baseline algorithms for several underlying graph based models.
Online Meeting Scheduler
- Worked in a team of six with scrum framework to build an online service to find optimal meeting time for large groups using Rails
- Scaled with server side caching and optimized database queries to handle 50k+ users/min.
Code - Report - Presentation
Electricity Generation Prediction Model (Jan-May’19 - In collaboration with govt. of India)
- Built and compared multiple generation prediction models (time series analysis).
- Yielded over 75% accuracy with 5 years of training data
Python Garbage Collector Analysis
- Understood internal functionality and parameters for Python and PyPy GC.
- Created custom visualizer to examine performance effects with parameters.
- Compared performance with state-of-the-art GC algorithms.
Blockchain Model
- Developed and implemented a distributed consensus protocol (algorithm) for distributed ledger (Blockchain).
- 1000+ lines of code in Python with socket programming and multiprocessing.
Anonymous Elections on Blockchains
- Developed and implemented anonymous voting protocol to conceal the voter identities using zero-knowledge proofs.
- Python implementation of the protocol scaled efficiently on a private ethereum network with only a linear time overhead compared to the conspicuous protocol.
Code - Report - Presentation