Jan 2024 – Jul 2024 · 6 mos · Bengaluru, Karnataka, India · On-site

• Developed LightGBM Models: Engineered and deployed LightGBM models in Python to accurately predict payout per click for each ad request, enhancing revenue predictions.
• Data Pipeline Creation: Designed and implemented robust data pipelines using Spark and Hadoop, efficiently filtering and processing terabytes of training data in Java.
• API Integration and Optimization: Integrated predictive models into the API serving code, optimizing performance to handle millions of requests per day with millisecond-level latency.
• Performance Enhancement: Conducted extensive optimizations to ensure low-latency responses and high throughput, improving the system's ability to manage high traffic volumes, by migrating to Aerospike .
• Cross-Functional Collaboration: Collaborated closely with data engineers, software developers, and product teams to ensure seamless integration and deployment of machine learning solutions.

Jul 2022 – Jan 2024 · 1 yr 6 mos · Bengaluru, Karnataka, India · On-site

• Working on developing a high frequency Trans-Atlantic arbitrage strategy for stocks listed in Europe and US.
• Also worked on a high-frequency cryptocurrency trading platform that leveraged arbitrage opportunities between different exchanges.
• Spearheaded the architecture and implementation of trading bots that executed trades identified by the arbitrage identifier.
• Utilized C++, Python and Golang to design and develop microservices for analyzing arbitrage opportunities and executing trades on corresponding exchanges, streamlining the trading process and reducing latency by 80% through the migration from REST API to Financial Information Exchange protocol.

Jan 2022 – Jul 2022 · 6 mos · Bangalore Urban, Karnataka, India

• Worked in the Nutanix Cluster Checker team, which is responsible to check the health of clusters and raise alerts to users if services fail.
• Worked on distributed log collection service which is responsible for collecting service logs which help in debugging.
• Generated automatic documentation for python code and hosted that on a website using Sphinx and Nginx.

Aug 2021 – Dec 2021 · 4 mos

• Used large scale pretrained model (CLIP) for few-shot, multilabel, and continual learning tasks.
• Developed a technique, called Multi-label Weight Imprinting, which uses the encoded image from the pretrained model to offer personalised labels.
• Worked on developing and testing an app used for AI-guided habit formation in a team of 5 students.
• Under Prof. Pattie Maes and Mina Khan

May 2021 – Jul 2021 · 2 mos · Mumbai, Maharashtra, India

• Worked on an end-to-end Attention-based model architecture to predict whether the customer will buy a product displayed in an Ad.
• Benchmarked the model with models in production.
• Developed a model which gave a boost of 36% in performance and 28% faster predictions compared to the model in production.

Jan 2021 – Jun 2021 · 5 mos · California, United States · Remote

• Worked on simulating and analysing the neural behaviour in dynamical models under different noise structures
• Used a vector-autoregressive model to model the neural behaviour. Added multiple models to vary the generation of matrices responsible for controlling the neural behaviour.
• Wrote SLURM scripts for running experiments on a high performance computer (NERSC) and integrated them with OpenMPI (using mpi4py) to parallelise python code.
• Was under the supervision of Professor Kristofer Bouchard at Neural Systems and Datascience Laboratory, Lawrence Berkeley National Laboratory.

Jul 2019 – Jul 2020 · 1 yr · Goa

• Head of the Back-end Team which manages the Back-end of the Student Welfare Division website

Jan 2019 – Jul 2019 · 6 mos · Goa

• Collaborator of the website development team of Student Welfare Division (SWD) Bits Goa. This website is built on Django

May 2019 – Jul 2019 · 2 mos · New Delhi Area, India

• Worked on a project called Rapid CT which enables radiologists to quickly identify whether a patient needs critical attention and should be added to a priority queue which is maintained at the hospital. This was done through deep learning. The CT scan of a patient would be the input for the model and the model was able to determine whether the patient has hemorrhage or not with a 98.3% accuracy. This significantly reduced the workload of the radiologist and helped him achieve results faster.
• The second project includes detecting the heart and plotting the heartbeat of a Zebrafish through a video of the fish.

Jan 2019 – May 2019 · 4 mos · Goa

• Teaching assistant of Computer Programming

Jul 2018 – May 2019 · 10 mos

May 2017 – Sep 2018 · 1 yr 4 mos