Oct 2020 – May 2022 · 1 yr 7 mos · Bengaluru, Karnataka, India

• Worked in a team of 3 to deliver Aspect-based Sentiment Analysis(ABSA) service - one of the 5 Natural Language services that were globally launched as part of OCI’s first AI service.
• Used horovod and slurm clusters to accelerate training of Natural Language models, slashing the training time five-folds.
• Developed a data augmentation service using GAN to generate fully tagged open/specific domain training data for language models.
• Developed a CI/CD pipeline for model training, tracking its metrics, also evaluating and serving the model using Nvidia Triton. This helped boost the inference performance of the model and reduced the development-deployment cycle by a good margin.

Jul 2019 – Sep 2020 · 1 yr 2 mos · Bangalore

• Worked on distributed deep learning models. Performed data parallelism using dask and model parallelism using all-ring reduce. Eliminated memory bottlenecks and model training time dipped by around 30 percent
• Working on improving recommendation engines and campaign personalisation models as well as their data pipeline architecture.
• Working on machine learning training and inference automation pipeline using Apache Spark, dask and MLFlow.
• Built an end to end pipeline to automatically deploy/train/test machine learning models on a hadoop cluster using apache airflow. Reduced 2 days of human effort to a REST call.

May 2018 – Jul 2018 · 2 mos · Bengaluru, Karnataka, India

• Worked on Apache Spark cluster optimization and brought the launch time from 22 minutes to an average of 3 minutes resulting in an 80% dip in costs.
• Worked on Fashion AI to recommend best-suited merchandise for the person according to his/her looks. Mapped the body demo-graphics to the suitable merchandise features using Inception-V3.
• Optimized the inference parts and the results for trained models on Spark using Apache Livy.
• Worked on StoreCare API on nodejs to make customer-sales integration.

Dec 2017 – Dec 2018 · 1 yr

• The Department of Visual Media (DVM) is a student body that is responsible for all the technical support in organizing the cultural and technical festivals at BITS Pilani. This includes the development of Website, Payment Gateway and the Registration Software along with coverage of the fest and making of teasers and inaugural videos.
• My work involves providing supervision and technical assistance to the team of 50+ undergraduates in carrying out the above functions smoothly

Aug 2015 – Oct 2018 · 3 yrs 2 mos

Dec 2017 – Dec 2018 · 1 yr

• One of the six members of the body managing India's largest completely student-organized cultural festival, Oasis. Oasis, with 45 years of its existence, attracts close to 3000 participants from over 100 colleges in India and has a budget upwards of 1.5 Cr INR. The StuCCA body takes all the decisions relating to the festival ranging from financial intricacies to operational details.

Aug 2017 – Jun 2018 · 10 mos · Pilani

• Students' Union Technical team of my University is responsible for the working of the institutes ERP system. As part of the duties of the Head of Technical team I am in-charge of web applications development and software development of the university.‌

May 2017 – Sep 2017 · 4 mos

• Working with P2PSP organisation on project of virtual room developing an application layer to implement the concept of many to many peer real data streaming

Mar 2017 – Sep 2018 · 1 yr 6 mos

• Quickly downloads, clean up, and install ecological datasets into a database management system.
• I added multiple datasets, added many important features such as automated credentials lookup, improved encodings of databases
• and currently working on adding spatial datasets.
• Helped in removing cyclic redundancy of generation of python and scripts

May 2016 – Nov 2017 · 1 yr 6 mos · Greater Delhi Area

• Worked on the backend services and prediction algorithms for a seed-funding Fintech Startup.
• The initial version of the product was built on Python/Django for beta-testing.
• Scraping Mutual Funds ticker data into PostGreSQL DB and manipulating the data to make frequency and user fund graphs separately.
• Predicting the financial health of each user by using momentum based strategy on mutual fund market
• Implemented a cart-parser for mutual funds and one-stop payment.