About
I'm a certified TensorFlow ML Engineer, and have publications at the SciPyConf. Read my latest research - here
Work Experience
Skills
Check out my latest work
I've worked on a variety of projects, from simple websites to complex web applications. Here are a few of my favorites.
Transtokenization
Adapted Mistral-7B LLM from English to Hindi through innovative tokenizer reconfiguration and cross-lingual token mapping. Implemented subword unit alignment for better coverage and fine-tuned with Hindi data, achieving significant improvements in perplexity, BLEU, and CHRF scores without full retraining.
AI Agent based Pull Request Reviewer
An autonomous code review agent system that uses AI to analyze GitHub pull requests. Made with Celery worker and Redis as broker for efficient API handling. It is based on CrewAI on top of OpenAI API that receives , analyzes code and outputs responses in JSON format.
Digital Divide AI
The digitaldivide.ai project is a collaborative initiative between XRI Global and students at the University of Arizona. Our mission is to provide real-time visibility into AI model capabilities across the world's languages. We believe that bridging the digital divide begins with establishing foundational technology support for all languages.
AI based changelog
A developer-facing tool that allows devs to quickly AI-generate a changelog. Provides a unique flow with github actions, github api, openai api, and json files to help generate changelog automatically on publish a release, also keeping the option to add things manually.
Side Quests
I have written and publshed research papers, got scholarships from Hack The North and Linux foundation to attend Kubecon. Below are some of my achievements listed.
- H
Hack The North
Waterloo, Ontario
I got a scholarship to attend Hack the North, Canada's largest hackathon organized by University of Waterloo - D
Delivered a speech around entrepreneurship.
Kolkata, India
I talked about how we raised funds for our startup with the initial MVP and scaled it for customers. https://www.linkedin.com/posts/academy-of-technology_a-special-talk-on-how-to-build-prize-winning-activity-7184774748038467584-P74E?utm_source=share&utm_medium=member_desktop&rcm=ACoAADb5VDMBBWBtC1aB-Mu0CuCEosyNC9Wd73g - Y
Yet another talk around machine learning operations on cloud-native technologies.
Kolkata, India
Here I discussed around machine learning operations creating entire pipelines of machine learning using Kubeflow and talking about an end-to-end ML Ops project. - T
Tensorflow Certfication
- S
SciPy Conference
Tacoma, WA, USA
My paper was accepted in this conference - S
Speaker at FOSS Kolkata
Kolkata, India
I delivered a speech around trans tokenization on large language models (monolingual). I spoke about fine-tuning adapting trans tokenization to your existing methods. Also talked about making your individual tokenizer. (https://www.linkedin.com/feed/update/urn:li:activity:7248060113524031488/) - K
Kubecon India
Delhi, India
Got a scholarship to attend Kubecon India - C
Conducted an AI session around trans-tokenization
Remote, India
Trans-tokenization topic.
Get in Touch
Want to chat? Just shoot me a dm with a direct question on twitter and I'll respond whenever I can. I will ignore all soliciting.