HOOMAAN SAJJADI

As an experienced Applied Machine Learning Engineer and Software Engineer, I specialize in healthcare applications and real-time data platforms. My passion for healthcare technology was sparked during my undergraduate studies and has been further developed through advanced study in Computer Science at Texas A&M University and my current role at a healthcare startup. I've expanded my expertise to include consumer medical software, adding another layer of versatility to my skill set. I am open to opportunities where I can further apply my skills and make a significant impact.

 Sudoku Solver Desktop App

• Developed a C++ graphical desktop application capable of solving advanced Sudoku puzzles. The application could accept a Sudoku table at any state and either solve it completely or declare it unsolvable.

3D Solar System Simulation

• Used OpenGL library and CUDA programming to create a 3D simulation of the solar system.

Blood Pressure Estimation

• Developed a multitask LSTM model for estimating blood pressure from bio-impedance signals. The tasks were systolic and diastolic blood pressures. Added auxiliary input of the first derivative of the signals and some preprocessing techniques to improve the results.

Meal Macronutrient Estimation

• Automated diet monitoring by predicting meal macronutrient content using continuous glucose monitors and deep learning. Explored several different models such as tree-based models and deep learning end-to-end models. Proposed a novel normalization technique to mitigate the inter-subject discrepancy in the data, which improved the accuracy of the model drastically.

Arithmetic Calculator with AVR Microprocessors

• Coded a software that gets an arithmetic operation as string input and analyzes its parts. Structurally and systematically solved and printed the results. Implemented the software to an actual AVR device using keyboard and LCD.

Marketplace Web App

• Developed a Python-based marketplace web app leveraging Django library and SQL database. The app simulated a selling and buying website with seller pages where they could put items in and the number of items available. The app also had a customer side for choosing which item they want and how much money they have to spend on purchasing the items.

Viterbi Algorithm GPU Parallelization

• Used CUDA programming language to implement the Viterbi algorithm in a GPU-optimized way. The algorithm was performing faster due to the single instruction multiple data fashion of the GPU.

Chatbot Web App with OpenAI GPT-4 API

• Built a web UI for real-world chatbot simulation using Python, Flask, and OpenAI GPT-4 API. Started from an interactive chat module and gradually improved its elements and features to make it look more like a personalized chat bot.

Real-time Sensor Data Collection App

• Developed mobile apps with watch companions on both Android and iOS platforms for real-time collection of sensor data from smartwatches and phones. The apps collected data from various sensors, including accelerometer, gyroscope, heart rate monitor, and more. The data was stored on the device and could be transferred to the phone when connected. The apps were designed to handle different data frequencies and ensure reliable data delivery even in case of connection failures.

• Improved the Mean Absolute Percentage Error by around 40% on average across macronutrients (protein, fat, carbs) for subject-independent models in the automatic diet monitoring project using subject normalization preprocessing techniques and accounting for fasting glucose
• Improved the accuracy of mortality prediction by more than 5% using advanced parameterization methods on electronic health record outcome prediction models.