Projects
Phi AI for Low/No Code (Pega) - Aug-2024
Phi AI for Pega is a Chrome extension that leverages two key functionalities: it analyzes critical business logic in Pega, identifying and resolving issues in activities and data transforms to prevent production failures, and it also uses Retrieval-Augmented Generation (RAG) with Pinecone to answer Pega domain-specific questions. This combination enhances development efficiency by streamlining workflows, improving code quality, and offering instant insights.
Trading Agent using Reinforcement Learning - June-2024
This project works on developing a Deep Reinforcement Learning (DRL) trading agent designed for continuous action spaces. We focus on algorithms like Deep Q-Network (DQN) and Actor-Critic, which are well-suited for such environments. Additionally, we design optimized reward functions and state representations to capture key financial metrics and market dynamics, enabling the agent to learn effective trading strategies.
Accelarating general-purpose applications using NN - March-2024
Parrot Transformation implements a neural network-based approach to accelerate general-purpose grayscale transformation functions. By replacing traditional methods with a neural network, the project achieves up to a 30% increase in inference speed using GPU and TPU accelerators. The model demonstrates high accuracy (90%) in emulating the original function, showcasing its effectiveness in diverse applications and emphasizing the practical benefits of hardware acceleration.
