Projects | Shree Venkatesh

Guided Payload

The guided payload is crucial to our macro UAV project. It features an autonomous self-guided TPU housing with a small water bottle inside, released mid-air from under the UAV's wings.

It is powered by a 2S Li-Po battery, an ESP32 microcontroller, a BN880 GPS/Compass module, 2 servos, and a Proportional Control Algorithm. The housing is attached to the parachute with strings, and steering strings connect to a custom 3D-printed servo arm on either side.

The servo pulls on the strings to steer the parachute, ensuring the payload is safely and accurately dropped onto the target.

Houston : Ground Control Station Frontend

The frontend of our ground control station, Houston, is a React web app for monitoring and adjusting plane parameters during flight. It tracks connectivity, telemetry (altitude, battery voltages, airspeed), and image data, allows in-flight camera adjustments, and facilitates mission parameter input and waypoint uploads. I designed and developed the UI for the Controls Page and the Report Page, and integrated the backend for the Report Page.

Stock Predictor

I developed a CNN using Python and TensorFlow to predict short-term stock behavior with 71% accuracy.

I preprocessed and normalized historical price data, used convolutional layers to extract temporal patterns, and tuned hyperparameters with the Keras Hyperband Tuner. The model was integrated into a backtested trading system for automated decision-making.

Snake AI

This project uses reinforcement learning to enable the model to learn from its environment and improve its performance over time. I set the agent to be the snake, the positive reward to be the food, the negative reward to be hitting itself, and the environment to be the playable space.

The model's ability to learn and adapt led to increasingly efficient gameplay, showcasing the potential of reinforcement learning in developing intelligent and adaptive systems. This project not only highlighted the versatility of PyTorch but also illustrated the practical applications of reinforcement learning in gaming and beyond.

Personal Portfolio Website

Built using HTML, CSS, JavaScript, and Jekyll, hosted on GitHub Pages. Designed and built from scratch, with a focus on clean UI and mobile responsiveness — the same design sensibility reflected in projects like Houston.

CSE 151A: NBA Odds Analysis

A group project for CSE 151A (Intro to ML) predicting NBA betting metrics — money lines, totals, spreads, and scores — using NN, RNN, and CNN models trained on historical NBA data.

The models surfaced how public sentiment toward team strengths influences betting odds, offering data-driven insight into which teams are favored and by how much.

DECaF 2024 Website: decafucsd.com

As the Co-President of Triton Engineering Student Council(TESC), I coordinated logistics for the Disciplines of Engineering Career Fair (DECaF), reaching 350+ attendees providing career opportunities for students. I designed the UI, and developed the website for DECaF from scratch using Figma, and React TS.

Triton UAS

Personal

Academic