Work Experience

A summary of my professional experience—from biotech and medical imaging to ML research and data science internships.

Mycrospace — Biotech Startup

Formulated the high-density colony counting problem (>300 overlapping colonies per plate) as a density estimation task and designed proprietary Density Map Regressors with Bayesian priors.
Designed and trained Object Detection models, achieving 95% median accuracy against expert microbiologist annotations.
Built complete production system end-to-end: PyTorch training framework, data pipelines, containerized inference (Docker/GCP), REST APIs (NestJS/Prisma), and user-facing web application (Next.js/TypeScript).
Curated domain-specific dataset of 3,500 annotated images across 12 microbial species; co-designed annotation protocols with microbiologists to ensure ground-truth integrity.
Delivered full-stack ML product from concept to deployment, owning model design, training infrastructure, API development, database architecture, and frontend implementation.

Trained and evaluated nnU-Net models for volumetric brain MRI segmentation, experimenting with different multi-scale feature extraction strategies.
Contributed to optimizing the inference pipeline for the VolBrain platform (which contains 700k+ MRI volumes), analyzing bottlenecks in data flow from NIfTI ingestion to voxel-level predictions.
Established systematic experiment tracking and modular code architecture, enabling reproducible ablation studies and standardized evaluation for the research group.

Built ML models for temporal gaze-pattern analysis: extracted features from raw ocular movement sequences, designed predictive pipelines, and evaluated against behavioural ground truth.

Early exposure to large language models (GPT-2/3) and Diffusion Models; contributed to applied NLP and generative modelling projects.