Supercharging Ramp with Applied AI

I am a CS student at NYU Courant. Interested in ML, NLP, Compilers and Graphics.

1. Branches: Built a framework to visualize advanced reasoning and planning algorithms like Tree Search with LLMs. Worked on

Tree-of-Thought + Reflexion style planning algorithms for objective problem solving.

2. Verilog Code Gen: Explored LLMs for Verilog Code Generation, using prompting, contrained generation, execution feedback and

reinforcement learning to solve the VerilogEval dataset.

I primarily worked on ML Automation of Taxonomy Tagging at Scale. I wore many hats like Data Scientist and Machine Learning Engineer.

1. Developed and Deployed over 250+ image models for various attributes like sleeve_length, pattern, etc for over 35 product categories

2. Color quantization models using Vision Transformers and K-Means Clustering

3. A baseline brute force continual learning framework for periodic retraining of models.

4. Model tracking and experimentation using neptune.ai

5. Deployment of batch-mode inference using Pyspark + EMR + Airflow

6. Python project development using nbdev.fast.ai

Apart from technical work, I proactively:

1. Hosted a weekly research & talks forum

2. Mentored juniors and interns

3. Wrote blog posts

4. Gave work demos and presented in org-wide forums

A Research Intern with the "Vision for Mobility" Group under CVIT in IIIT-H. Did some research on:

1. Occluded Object Detection of Traffic Signs.

2. Missing object detection