Tejas Khare

Data Scientist @Advanced Sensor Research

Ottawa, ON, CA

Researching on Causal Radio-Frequency & Optics applications in non-invasive CGM.Architected various models - LSTM composite autoencoders, GNN, and CNN for non-invasive glucose measurement via adaptive, calibration-aware sensor fusion.Advanced translational machine learning for NIRS, leveraging manifold learning and latent space analysis to surface clinically actionable glucose biomarkers.Drove robust model generalization and class balance using SMOGN data augmentation and context-driven sequential modeling for real-world glucose diagnostics.Orchestrated end-to-end ML pipelines for production, integrating Google Cloud workflows and enforcing automated data QA to enable regulatory-grade deployment.Partnered with regulatory and clinical teams to ensure model compliance (FDA/CE-MDR) and to translate ambiguous ‘patient experience’ goals into concrete data collection, labeling, and model‑evaluation proceduresDesigned evaluation pipelines that connected sensor‑level metrics (NIR, PPG, accelerometer) to clinically meaningful outcomes for people with diabetes, iterating on metrics to better reflect real‑world utility.Used SHAP, LIME, and Grad‑CAM to study where models disagreed with human experts, treating explanations and error patterns as feedback signals to improve both features and architectures.ASR (A part of Medwatch Technologies Inc.)

University of Ottawa

Master's degree, Computer Science Applied Artificial Intelligence

MIT World Peace University

Bachelor of Technology - BTech, Electronics and Communications Engineering