Animesh Jha

Hi! I am Animesh, I recently graduated from the Department of Computer Science and Engineering at the Indian Institute of Technology, Kharagpur.

Presently, I am working as a Software Engineer at Rubrik.

I have had a lot of fun exploring Theoretical Machine Learning, Federated Learning, Security, and Robotics through my various experiences. I enjoy learning about all kinds of random new things and was an active member of the Quiz Club at IIT KGP.

Feel free to check out my resume and drop me an e-mail if you want to chat with me!

Email  /  CV (March 2023)  /  Github  /  LinkedIn  /  Google Scholar

profile photo
News

  • [June '23]  Joined full time at Rubrik Bangalore
  • [May '23]  Graduated from IIT Kharagpur :)
  • [Feb '23]  Captained the team of IIT Kharagpur and won Silver in the DevRev Efficient Domain Specific QA Event at Inter IIT Tech Meet 2023
  • [Jan '23]  Member of the Silver winning Quiz Team of IIT Kharagpur at Nihilanth 2023
  • [Jan '23]  Member of the Silver winning Quiz Team of IIT Kharagpur in 5th Inter IIT Cultural Meet 2023
  • [March '22]  Our team won Gold in the Bosch Model Extraction Event at Inter IIT Tech Meet 2022
  • [August '21]  Our team (NaN) qualified for and placed 44th at the Gwalior Pune ICPC Regionals 2020
  • [May '21]  Our work on controls of Autonomous Racing was accepted at the OCAR Workshop ICRA 2021
  • [March '21]  Our team (NaN) was amongst the top 5 National Finalists at the Uber Hacktag 2021
  • [March '21]  Participated in the Automatic Headline and Sentiment Generator Event at Inter IIT Tech Meet 2021
  • [March '20]  Joined Autonomous Ground Vehicle (AGV) IIT KGP as a member of the Software and AI Team
  • [December '19]  Member of the Gold winning Quiz Team of IIT Kharagpur in 4th Inter IIT Cultural Meet 2019
Publications
Vroom Vroom Local NMPC on Global Optimised Path for Autonomous Racing
Dvij Kalaria*, Parv Maheshwari*, Animesh Jha*, Arnesh Kumar Issar*, Debashish Chakravarty, Sohel Anwar, Andres Tovar,   OCAR Workshop, ICRA 2021
paper | code

We present a novel strategy for the control of an autonomous racing car on a pre-mapped track. Using a dynamic model of the vehicle, the optimal racing line is computed. The tire forces at high speeds are modelled using a modified Pacejka model. These are then used with a local Nonlinear Model Predictive Control which accounts for drafting in overtaking scenarios and merges with the reference line at turns to make greater progress.
RE DSPT [Re]: Differentiable Spatial Planning using Transformers
Rohit Ranjan*, Himadri Bhakta*, Animesh Jha, Parv Maheshwari   MLRC 2021, ReScience C, NeurIPS Journal Track 2022
paper
RE DSPT [Re]: Contrastive Learning of Socially-aware Motion Representations
Roopsa Sen, Sidharth Sinha, Parv Maheshwari, Animesh Jha   MLRC 2021, ReScience C, NeurIPS Journal Track 2022
paper
Projects
Efficient and Provably Byzantine Robust Optimization  
Undergraduate Thesis Project (Advised by Prof. Simon S. Du and Prof. Swagato Sanyal)

Designed and analysed Byzantine Fault Tolerant algorithms for efficient Distributed Non Convex First Order Optimisation. Established lower bounds on the number of iterations for finding ϵ approximate saddle points in the presence of adversaries. Work submitted at AISTATS and was awarded Best Undergraduate Thesis Award by CSE, IIT Kharagpur.

Apache Spark on Kubernetes  
Internship Project, Rubrik India

Migrated datapipelines handling 10+ TB per day from AWS EMR to Kubernetes using the Spark Operator for Kubernetes. Setup observability and monitoring for production deployment, pushed container logs to CloudWatch and S3 via FluentBit. Collected job and cluster metrics via Prometheus, created Grafana dashboards to analyse and improve resource utilisation. Implemented cluster autoscaling to dynamically adjust resources for low utilization periods. Identified optimal node sizes for efficient bin packing to reduce compute costs. Leveraged AWS Spot nodes for running workers to maximize cost savings
Resource Efficient Domain Specific QA   [ code  |  report ]
Team Captain - Silver - Inter IIT Tech Meet 2023

Integrated sentence level context retrieval with an ensemble of noisy tuned LLMs with contrastive loss to extract answers. Achieved low latency via ONNX, Caching and Quantization. Experimented with MAML for efficient domain adaptation

Model Extraction for Video Transformers   [ code  ]
Gold - Inter IIT Tech Meet 2022

Performed model extraction attack on SwimT and Movinet in greybox setting by a MARS model trained on augmented data. Applied Data Free Model Extraction with CGAN and adversarially generated synthetic examples through incremental perturbation in the blackbox setting
Communication Efficient Federated Learning  
With Dr. Jihong Park

Improved communication efficiency of Federated Learning by reducing model size via sparsification through lottery tickets. Used supermasks to prune server-side models, reducing model size while maintaining accuracy and client-side data privacy.
Trajectory Generation in Frenet Frame  [ code  |  demo ]
AGV Planning Project

Open Source Python and C++ implementation for a sampling based planner in Frenet Frame. ROS1 and ROS2 support provided, with Gazebo and LGSVL support for simulations. OpenMP used for increasing the sampling speed of the planner.

Online Sales Portal  [ code ]
Software Engineering Lab Term Project

Created a Web App using Flask which provides features like notifications, categorization, upload and search of items.
Used an Object Document Mapper (MongoEngine) to store website data on a MongoDB database.

Discord Bot for Quizzing  [ code ]
Personal Project

Created Discord bot to make online quizzing easier on Discord. Helped the Quiz Club to continue during the pandemic. Used Discord.JS to implement features like scoreboard and buzzers and to automate delivery of messages to individuals. Used by nearly 100 people, multiple times.

ESPN Cricinfo from the Terminal  [ code ]
Personal Project

Developed a bot which fetches player data and live scores from ESPNCricinfo and displays it on the terminal.
Can parse complex player statistics queries, retrieve it from ESPN statsguru and display it in a readable format.
Not really sure of the point of the project though.


Design and source code from Jon Barron