Aiden Robinson

Spincard

• Developing a mobile app that allows small groups to create virtual credit cards and split expenses automatically
• A serverless, microservice-based platform for pooling and spending money. I worked on backend development, including database setup, migrations, GraphQL server, and cloud deployment
• Built with Go, GraphQL, PostgreSQL, Clerk, Encore, and React Native

Repo

AR.cade, Hack Western Winner

• Uses OpenCV and Google Mediapipe to detect user movements, allowing players to control classic games like Temple Run and Super Mario through body gestures.
• Developed a custom body-part movement algorithm to map gestures from detected positions (e.g., raising an arm) to specific in-game actions
• Tackled embedding real-time video with body tracking overlay into the frontend using API calls, managing latency with threading optimizations
• Placed 3rd overall of 70 total projects

Devpost

Locker as a Service, DeltaHacks Winner

• Provides a secure, IoT-enabled food locker solution controlled via a mobile app, allowing remote lock/unlock functionality with biometric access and temperature control for safe, contactless food delivery
• Built with a Raspberry Pi, Firebase, Python, C++, and Flutter, utilizing real-time IoT data to control locker functions, including a servo motor and temperature settings
• Placed 3rd overall of 95 total projects

Devpost

Computer Vision Touchless Keyboard

• Developed a touchless keyboard that uses computer vision to detect finger strokes and plays musical notes via a buzzer controlled by an Arduino
• Utilized Google’s MediaPipe framework for hand tracking with 21 landmarks and OpenCV for capturing webcam data and visual feedback. Landmark coordinates were converted into screen ratios to accurately detect finger positions
• Established serial communication between Python and Arduino, sending note data from Python to trigger corresponding buzzer frequencies on the Arduino

Repo

Embedded Pacemaker

• Created a Hardware in the Loop and Software in the Loop pacemaker to represent dynamic pacing modes of a pacemaker using accelerometer data on an STM32 microcontroller
• Data framing, data processing, and stateflows were implemented in Simulink
• Supports 8 pacing modes: (AOO,VOO,AAI,VVI,AOOR,VOOR,AAIR,VVIR,DDDR)
• Frontend created in Python and connects via UART protocol, including a real time electrocardiogram

Repo

Segment Display Neural Network

• Developed a segment display recognition neural network from scratch in C++ that identifies digits 0-9 based on varying brightness and incomplete segments
• The model is a fully connected feedforward network with an input layer, hidden layer, and output layer, using sigmoid activation functions. The 7 inputs represent segment brightness levels, and 10 outputs indicate confidence levels for each digit
• Trained using backpropagation for 3,000 epochs with an initial set of ideal patterns, achieving a training error of 0.003

Repo

Second Brain For Students, CUCAI '24

• Led a team of 6 to develop "SecondBrain," an NLP-based AI assistant to help students manage and query personal and academic documents efficiently
• Developed with LangChain, Pinecone VDB, React.js, Flask, OpenAI API – Users upload documents (PDF/Docx), which are processed, vectorized, and stored in a unique database. When users ask questions, the system retrieves relevant sections using similarity search and GPT-4 for clear answers
• Reduced token costs by 45% through efficient data management and API integration

Repo

Sentiment Analysis Neural Network

• Built a neural network model to classify IMDb movie reviews as positive or negative, using a dataset of 25,000 polar reviews
• Utilized TensorFlow for training a sequential neural network. The reviews were preprocessed (standardized to 250 words, vocabulary size set to 88,000) and encoded into integer sequences using a custom word index
• Model architecture consists of embedding layer, global average pooling layer, dense layer, and output layer
• Trained on a split dataset (90% training, 10% validation) for 40 epochs, achieving accuracy of 99.9% on a personal review.

Repo

NLP Keyword Context Tracker

• Developed a tool to search PDFs by keywords, clustering their meanings to return the top 3 relevant paragraphs, enhancing document searchability
• Using Python, Cohere NLP API, Annoy, Pandas, React.js, Flask – The user inputs a PDF file and keywords. The tool preprocesses the data, uses semantic search to compare keyword vectors with paragraph embeddings, and returns the most relevant paragraphs with page numbers

Repo

Web Scraper for Collectible Prices

• Developed an automated web scraper to track the prices of collectible items (e.g., Pokémon cards) across various websites, saving an hour of manual work weekly
• The script fetches HTML from URLs, isolates price data using Beautiful Soup, processes it into structured tables, and sorts the data by price using Pandas for easy viewing

Repo

High Voltage Controller Respin

• Redesigned the high voltage safety controller, meeting Formula SAE Electric (FSAE) standards through iterative improvements based on competition feedback
• Led schematic design, PCB routing, part sourcing, manufacturing, and testing
• Enhanced circuit modularity for easier testing and integrated LED indicators for diagnostic clarity

PWM Relay Economizer

• Designed and manufactured relay economizer board using Altium Designer to be used on contactors throughout the car
• Optimizes solenoid power consumption by utilizing Pulse Width Modulation (PWM) to supply just enough current to keep contactors closed rather than continuously sending current in
• Measured activation and hold phases of components and chose components accordingly

Sequential Logic Design, 2E04 Project

• Designed and built a digital circuit using J-K flip-flops to display my student number on a 7-segment display, with a custom solution for tracking repeated digits in the binary sequence
• Created optimized truth tables and Karnaugh maps (K-maps) to derive logic expressions for each J-K flip-flop input, resulting in both SOP and POS logic models
• Simulated the final circuit using NI Multisim, achieving a clean timing diagram matching expected binary outputs for the display
• Assembled and tested the physical circuit with organized, color-coded wiring for clarity, successfully achieving real-time display of my student number