- B.Tech, Electrical Engineering; National Institute of Technology, 8.31/10.0
Senior Product Engineer, Sprinklr (2019 - Present)
Experience with training cutting edge models in NLP and Computer Vision and deploying them with k8s
Worked on training and deploying end-to-end OCR for Twitter Firehose. This was used by multiple clients to query over messages to reduce spam and take action on misinformed tweets.
Trained a token classification (NER) model to detect People, Brands and Products in tweets. We used this model to find the trending brands and products in social media conversations.
Software Intern, Morgan Stanley (2018)
Worked on Kafka, a horizontally scalable, fault-tolerant and fast messaging system, used for building real-time data pipelines and streaming apps. My project work included joining two streams of data using Kafka Streams API. The project was successfully deployed and tested within two months.
Member, Delta Force (2016 - 2017)
Worked on Laravel, a PHP framework, for deploying REST APIs responsible for handling all the college fests. Used by more than 4000 students from NIT Trichy and other colleges. As of 2021, this API is still being used by the club for all the major fests.
A Python based program to recognize the raag of a Carnatic piece. The program considers the tonal and temporal characteristics and uses the k-Nearest Neighbours technique find the raag of the music.
An image captioning program for blind people, trained on the Microsoft Coco Dataset with the “Show and Tell” model using TensorFlow. The model was run on a Raspberry Pi connected to a webcam. The program clicks an image, and reads out the caption of the image, with proper grammar.
The Beer Factory Game
A supply-chain management game, in which participants manage the production processes of a beer factory and compete against computer simulated opponents or fellow participants. The backend was coded in Django
A Python based command line application to download files matching given extension from any website. The package is available in pip as extscrape. It supports multithreaded downloading.
A real time fault monitoring system using Raspberry Pi and Intel Edison boards. The Raspberry and Edison act as nodes and measure voltage, current and send data to the server. The server stores the values in database. Administrators can login to the web app and monitor the values. The backend was written in Django framework.
- Programming Languages: Python, C++, Java
- Distributed: Kubernetes
- ML Frameworks: PyTorch, Tensorflow, Transformers (Huggingface)