AffordanceNet

Computer Vision

• Implemented an instance segmentation model to detect affordances of object parts in images, utilizing SegFormer (ViT based Transformer architecture) and two CNN decoders trained with a joint loss function.

InspiredResume

LLMs

• Product that leverages AI to help job applicants with customized resumes and cover letters based on the job description It also helps users format and remove grammatical errors from the resume.

GridWorld

Reinforcement Learning

• Optimized agent navigation in a grid environment using Proximal Policy Optimization (PPO) with Behavioral Cloning for enhanced learning efficiency.

Semi-Supervised Learning

Computer Vision

• Explored scalability of deep neural networks with limited labeled data using Semi-Supervised Learning techniques: Pseudo-label, Virtual Adversarial Training, and FixMatch.