Open to data scientist and ML scientist roles · Vancouver, Toronto, Montreal or remote

Gurleen Kaur

neuroscience × ML × clinical AI

I build machine learning for clinical data — VR movement biomarkers for early cognitive decline, computational neuropathology on post-mortem brain tissue, and multimodal models over CT imaging and electronic health records.

Email me Résumé (PDF) LinkedIn ↗ GitHub ↗

About

The human
behind the models.

I am a data scientist and ML researcher in Vancouver, BC. I completed my Master of Data Science at UBC in June 2026, and I work across three UBC research labs spanning biomedical engineering, neuropathology, and addiction psychiatry.

My background runs through electronics engineering, health informatics, and data science. That mix gives me the rigour to build ML pipelines that hold up and the domain knowledge to make them clinically meaningful.

Whether it is 6DOF VR sensor streams from patients at risk of Alzheimer's, post-mortem IHC brain slides, or large-scale health claims, I am drawn to data sitting at the edge of what medicine can currently see.

Outside the lab: hiking the North Shore, lifting heavy things.

Clinical ML Medical imaging Time series MDS, UBC 2026 Vancouver, BC
0.850 AUROC, capstone INSPECT: 12-month pulmonary embolism outcome from CTPA and EHR
84.5% Balanced accuracy VR cognitive decline classifier, 0.90 AUC, 45 participants
3 UBC research labs NC4, Hirsch-Reinshagen, ACD
🏆 Best oral presentation PATHS To A Cure Neuroscience Conference, UBC

Research

Active Research.

Three concurrent positions at UBC across cognitive neuroscience, neuropathology, and addiction psychiatry.

NC4 Lab — Neural Circuits for Computation, Cognition and Control Apr 2026 – Present

Cognitive decline detection from VR spatial navigation

Supervisor: Dr. Manu Madhav · UBC Biomedical Engineering

  • Built a time-series classification pipeline over 6DOF VR navigation streams — head orientation, avatar position, controller quaternions — from 45 participants, to detect early markers of cognitive decline.
  • A ROCKET-based classifier reached 84.5% balanced accuracy and 0.90 AUC under leave-one-participant-out cross-validation, so no participant appears in both training and test.
  • Functional PCA and Dynamic Time Warping for feature extraction on variable-length multivariate series, benchmarked against gradient-boosted models on handcrafted movement features such as angular velocity, heading drift, and rotation smoothness.
  • First-author manuscript in preparation.
ROCKETTime seriesLOPO CVfPCADTW6DOF VRAlzheimer's
Hirsch-Reinshagen Lab — Neuropathology and CNS Disorders Mar 2026 – Present

Computational IHC analysis for CNS disorders

Supervisor: Dr. Veronica Hirsch-Reinshagen · UBC Pathology / VGH

  • Image analysis of multiplexed IHC slides from post-mortem human brain tissue, using QuPath for cell segmentation and stain colocalisation.
  • ML pipelines for automated spatial profiling of protein markers across brain regions in neurodegeneration and traumatic spinal cord injury.
QuPathIHC analysisCell segmentationNeurodegenerationSpinal cord injury
ACD Research Group — Addictions and Concurrent Disorders Nov 2025 – Present

Opioid overdose, acquired brain injury and fentanyl clustering

Supervisor: Dr. Michael Krausz · UBC Psychiatry

  • Co-authoring a paper on opioid overdose and acquired brain injury: systematic review, data extraction, and manuscript writing.
  • Research on fentanyl overdose clustering and temporal patterns of overdose events across vulnerable populations.
  • Presented a systematic review on stimulant co-use in opioid agonist therapy at the PATHS To A Cure Neuroscience Conference, UBC.
Systematic reviewEpidemiologyOpioid ABIFentanyl clustering

🏆 Best Oral Presentation — PATHS To A Cure Neuroscience Conference, UBC

Skills

Technical Stack.

Grouped by how I actually use them.

Languages
PythonRSQLCBash
ML and AI
XGBoostLightGBMROCKETTransformersTime series
Libraries
PyTorchscikit-learnpandasOpenCVtslearn
Clinical and pathology
EHR/EMRIHCQuPathEpidemiologySystematic review
Tools and cloud
GitDockerGCPPostgreSQLMongoDB

Projects

Selected Projects.

Capstone and independent builds, separate from lab research.

Capstone 2026 · MDS at UBC · with Tandem Research

INSPECT: multimodal pulmonary embolism prediction

A late-fusion pipeline over Stanford's INSPECT benchmark, combining CTPA deep learning (ResNet encoder into a GRU sequence model) with structured EHR features (LightGBM) to predict pulmonary embolism and eight downstream clinical outcomes. 0.850 AUROC at 12 months and 0.809 at 6 months, delivered as a reproducible pipeline handed off to the industry partner. Trained on GCP with GCS storage.

PyTorchLightGBMGCPCTPAEHR
GitHub ↗

Drug discovery · Personal project

Molecular binding affinity with uncertainty

Predicts pChEMBL binding affinity against EGFR, a cancer-relevant kinase, over 13k ChEMBL molecules. Compares a Morgan fingerprint MLP against Graph Attention Networks, with MC Dropout uncertainty quantification on both and scaffold-split evaluation so held-out molecules are structurally distinct.

PyTorch GeometricGATMC DropoutOptuna
GitHub ↗

Personal project

BrainMRI-Net: Alzheimer's classification

Three-class MRI classification — cognitively normal, MCI, and Alzheimer's — on the OASIS-1 dataset. A CNN pipeline with Grad-CAM interpretability for radiologist-readable saliency, and subject-level splits so no participant's scans span both training and test.

PyTorchGrad-CAMOASIS-1CNNIn progress
GitHub ↗

Journey

Education and Experience.

Education Industry

  1. 2014 – Jul 2018

    BTech, Electronics & Telecom (AI)

    MIT Pune · India

    Signal processing, circuits, and an AI specialisation.

  2. Oct 2020 – Feb 2021

    Data Scientist

    KPMG India · New Delhi

    EMR/EHR analytics, heart disease prediction, and brain CT/MRI preprocessing pipelines.

  3. Mar 2021 – May 2022

    R&D Intern

    In-Med Prognostics · Pune

    Brain CT alignment pipeline for roughly 100 patients, standardising anatomical axes for downstream analysis.

  4. Jun 2022 – Jun 2025

    Data Scientist

    Ronsare Research · Remote

    Three years on healthcare claims ETL pipelines, ML preprocessing, and data governance workflows.

  5. Jan 2023 – Dec 2023

    Advanced Diploma, Technology Management

    RRC Polytech · Winnipeg

    Completed alongside remote work at Ronsare. Focus on data-driven decision making.

  6. Aug 2025 – Jun 2026

    Master of Data Science

    University of British Columbia

    Graduated June 2026. Capstone: INSPECT multimodal pulmonary embolism prediction with Tandem Research.

  7. Nov 2025 – Present

    Research Assistant × 3

    UBC · NC4, VGH, Psychiatry

    VR cognitive decline, IHC neuropathology, opioid overdose epidemiology. Best Oral Presentation award.

Publications and writing

Research and Writing.

Conference presentations, papers in progress, and technical writing.

Paper — in progress

Opioid overdose and acquired brain injury

Co-authored under Dr. Michael Krausz, UBC Psychiatry. Systematic review, data extraction, and manuscript writing on ABI prevalence in opioid-overdose populations.

In review

Paper — in preparation

VR spatial navigation markers of cognitive decline

First-author manuscript from the NC4 Lab on time-series classification of 6DOF navigation data, with leave-one-participant-out validation across 45 participants.

In preparation

Conference presentation · UBC

Stimulant co-use in opioid agonist therapy

Presented at the PATHS To A Cure Neuroscience Research Conference, UBC. 🏆 Best Oral Presentation.

2026

Technical writing · Medium

Articles on ML, clinical data science, and neuroscience

In-depth writing for a technical audience on modelling choices, validation, and what clinical data actually demands of a pipeline.

Read ↗

Let's build something meaningful.

I am open to data scientist and ML scientist roles in health and biotech. The fastest way to reach me is email.

gurleenlakhman@gmail.com LinkedIn ↗ GitHub ↗ Medium ↗