Projects

A real-time Formula 1 race strategy system combining pit stop probability prediction and optimal pit window timing to deliver a live frontend dashboard for real-time race strategy buildup with lowest possible latency.

A web app where musicians instantly get lessons and notes for any music piece and any known instrument through music transcription & AI suggestions via Onsets & Frames, Demucs and Basic Pitch, trained over 20000+ MP3 & MIDI files.

An online platform that connects writers to traditional publishers. Curated a data consisting of 626 publishing houses in the US and 17942 published authors to find trends that can best fit a budding writer match the suitable publishing house.

A web app that analyzes step-by-step math solutions from uploaded images, identifies errors, and provides feedback. Integrates Mathpix for LaTex conversion, WolframAlpha for analysis, and Imgur API for OCR, ChatGPT for feedback.

A web app that gives outfit recommendations and try-ons for your entire wardrobe using Ximilar Fashion Tagging model and Anthropic for generating trending fashion suggestions based on occasions.

Used the FRDM-KL46Z NXP microprocessor board with C to control a robot, to move in a figure-eight pattern, follow a line, avoid obstacles, and navigate through a colored maze, after configuring GPIO registors, interrupts, encoders and utilizing components' datasheets, PID tuning and I2C communication.

Developed a real-time alarm clock using Intel Quartus Lite and deployed the program to the Intel DE10-Lite FPGA board. Also developed an audio recorder that records and plays back up to eight seconds of audio in SystemVerilog in the Nexys A7-100 FPGA board via Xilinx Vivado.

A CNN based model built using Keras and Tensorflow for crop disease detection from a UAS. The 2D-CNN model uses 4 hidden layers using ReLu activation function with the output layer using Softmax function, achieving 89.23% accuracy.

Part of the team developing a Cubesat for NASA CSLI Launch 2024, establishing communication network between LoRa devices for greater coverage, range and penetration without current orbital store-and-forward methods.

Built an autonomous wall-maze and line-maze solving robot using Teensy, TB6612FNG motor driver, Polulu QTR, HCSR04 Ultrasonic Sensors, N20 motors and Power Distribution Circuit, connected via Bluetooth to a mobile app to calibrate PID values and switch between manual-to-autonomous drive as well as line-maze to wall-maze settings.