Mycrospace is a biotech startup I co-founded in early 2024. I lead as CTO, and together with my co-founder we're building the first AI-powered platform for microbiology labs: from plate to report.
Thousands of researchers still count colonies by hand. A single lab can process hundreds of plates per day—each with dozens or hundreds of colonies—and manual counting is slow, error-prone, and hard to scale. Data ends up on paper, reports are tedious to compile, and human error can compromise the quality of the research.
We built a full-stack product so labs can photograph a plate and get counts and analysis in seconds. The core challenge was high-density colony counting: when hundreds of colonies overlap on one plate, simple detection isn't enough. I formulated the problem as density estimation and designed proprietary Density Map Regressors with Bayesian priors, and separately trained Object Detection models that reach 95% median accuracy against expert microbiologist annotations.
On the engineering side, I own the whole pipeline: PyTorch training framework, data pipelines, containerized inference on GCP, REST APIs (NestJS, Prisma), and the user-facing web app (Next.js, TypeScript). We also curated a domain-specific dataset of 3,500 annotated images across 12 microbial species and co-designed annotation protocols with microbiologists to keep ground truth reliable.
In 2024 we won 1st Prize at the IdeasUPV Startup Competition, with a €10,000 grant for biotech innovation.
If you run or work in a microbiology lab and want to see how it could fit your workflow, you can explore the product and get in touch at mycrospace.es.