Relevant Courses: Machine Learning, Data Mining
Bachelor Thesis: Currently pursuing at School of Computer Science, CMU
Working on depth estimation and obstacle avoidance using vision for low-computational moving devices like Anki Cozmo.
Worked on Task-based Information Retrieval specifically recommendation by leveraging external ontology. Also worked on dynamic enrichment of ontology based on the philosophy of Bayesian Rose Trees.
Worked on adding more data repository support to DataDepsGenerators.jl, which inturn creates registration blocks for DataDeps.jl. DataDeps.jl helps in downloading scientific data at the touch of your fingertip.
Revamped the existing bug filing interface (enter bug.cgi) into a single page interface reducing the multi-hop db access, unnecessary template rendering etc. This enhances the speed and performance in bug filing.
- BITS Darshini: A Modular, Concurrent Protocol Analyzer Workbench
P. Talasila, M. Kakrambe S. Santy, A. Rai, N. Goveas, B. Deshpande
Proc. ACM ICDCN, 2018
- Spawned a cross-collaboration workbench for BITS Darshini, a modular, concurrent, cus- tomizable protocol analyzer. This is implemented for sharing the experiments among researchers and experts (for review) in order to save resources on analysis of large pcap files involved in the process.
- Developed a vagrant deployment of the application for cross-platform installation.
- Modularized and refactored BackboneJS based client module. Decoupled the persistence code from the rest of the codebase.
- Designing a proof-of-concept IoT scalable interface to gather data from sensors installed at the waste water treatment site. This work is in collaboration with Ohio University, Cranfield University, and Caltech.
- Used XbeeS2C which supports socket-based streaming and static data transfer.
- Further work involves powering the sensors (deployed on organic farm) using Microbial Fuel Cells.
- Developed a CLI for AutolabJS which helps in committing the code to local gitlab server, submit the code from gitlab server to the evaluation nodes and retrieve the results using a socket connection.
- Published on npm and currently in use for programming labs at the university.