Worked with EMG, ECG, point cloud, body markers from motion capture data as well as tabular data to train decision trees, support vector machines and random forest models using ScikitLearn and Python. I was responsible for data collection, data handling, statistical testing and ML model development. • Developed classical ML models such as decision trees, support vector machines as well as advanced methods such as message passing graph neural networks, representation learning as well as deep learning. • Worked on developing custom models for atrial fibrillation detection (92.1%), stress detection (MAE: 0.2 DAS values), automated filtering of data and 5 class action recognition (Accuracy 79%).

Developed an LLM application that generates reports from tabular data using OpenAI's GPT-3.5 Turbo, using multimodal retrieval-augmented generation (RAG) with crewAI and LangChain.

Created an optimization method that utilizes tabular and graph data to optimize scheduling for road-related tasks, implementing a Greedy Randomized Adaptive Search Procedure (GRASP).

Developed computer vision solutions including object detection for quality control using Python. Worked on object detection in images with OpenCV using YOLO, achieving 98% accuracy, and performed tasks such as image augmentation and labelling. • Created an image and video labelling tool and augmentation for image and video data using Python (PyTorch, Tensorflow, LabelImg and JAX) • Additionally, implemented human action recognition in videos by converting them into body landmarks using MPHolistic and creating a custom neural network classifier, which reached 88% validation accuracy.

• Conducted a comparison of machine learning models for predicting interstitial glucose using wearable data and food logs. • Processed and transformed time series data into features and events, which were used to develop graph neural networks (GNNs). Leveraged large language models (LLMs) to create a personalized healthcare assistant that provides patient insights based on healthcare data. • Additionally, developed new sleep-related features from healthcare data, resulting in an improvement in mean absolute error (MAE) from 10.01 mg/dL to 5.2 mg/dL.

• Developed and compared ML models to classify gestures using EMG sensors and IMU, to control an industrial robot

• Developed a control for upper limb prosthesis using Myo armband (EMG and IMU). I also developed haptic feedback for a trans humeral amputee using a rotating balance based on sensed weight using Matlab and Arduino