Fixed a production LLM memory bug that cut latency 150x. Built multi-agent AI for a global bank. LangGraph · Google ADK

Shreyan Pal  • Peerlist

A fully immersive 3D Portfolio Website! 🚀 Explore an island adventure while learning more about me, Shreyan Pal. Built with three.js, vite, and react.js, this site boasts stunning animations and full responsiveness. Don't miss the playful surprise waiting at the end!

3D Portfolio Website

CRUD STUD APP

DREADNOUGHT DASHBOARD

Co-built a full EdTech SaaS platform from zero — MySQL schema, Python REST APIs, React.js frontend with WebRTC live streaming, Auth0 authentication — serving 10,000+ active learners globally on AWS. Fixed the description typo ("up tj date" → "up to date") and remove the "100% placement guarantee" line — unverifiable claims in project descriptions damage credibility.

Ed- Tech Website | Droupout

Finger Print based Voting System

You can like movies as per your preferrence and get ratings and genres

IMDB WEBSITE (CLONE)

Lyrics Finder

Grand Finalist, TVS Motor Innovation Challenge 2024 (IIT Madras, 500+ teams). Built an AI-powered smart showroom system — Gemini NLP chatbot handling natural language queries, ultrasonic sensors, interactive displays, and voice interfaces. Designed to adapt to Gen-Z language patterns, misspellings, and short-form queries for a human-less dealership experience.

Mitra

Built a fully functional myoelectric prosthetic arm — EMG sensors capture nerve electrical signals, Arduino C/C++ processes them in real time, servo motors actuate across anatomically-researched joint angles. Complete hardware-software integration at biomedical scale. This project is why I think about latency differently: human intent → machine response is a problem I've been solving since before I wrote my first API.

Myoelectric Prosthetic Arm

NewsMonkey

portfolio website

Our project aims to develop a self-driving car using deep learning techniques and the Deep Q-Network (DQN) algorithm implemented in Python. The self-driving car will have the ability to learn new routes and adapt to different environments.To enhance the learning process, we will employ the DQN algorithm, which combines deep learning with reinforcement learning. The DQN agent will interact with the environment (road and traffic) and learn to make decisions based on the input images. It will receive rewards or penalties based on the car's performance and use these rewards to update its policy and improve its driving behavior over time.To enable the self-driving car to learn new routes, we will incorporate a route learning mechanism. When faced with a new route, the car will explore the environment, observe the road, and gather data. It will then use this data to update its knowledge and adapt its driving strategy. This process will enable the car to gradually learn and navigate new routes efficiently. The entire project will be implemented in Python, utilizing deep learning libraries such as TensorFlow or PyTorch for building and training the neural network models. We will also leverage OpenAI Gym or similar frameworks for implementing the DQN algorithm and reinforcement learning environment.

Self Driving Car

Made a smoke detector using Arduino Uno and a buzzer. As soon as the smoke detector detects the smoke the buzzer buzzes.