Space BioLab
A modular micro-gravity lab to make orbital biology accessible.
Space BioLab is a service that opens the doors to orbital labs — a 16U CubeSat on the NanoAvionics M16P bus, divided into modular cartridges that universities, biotech startups, and pharma companies rent to run autonomous biomedical experiments in microgravity.
The 2025 NASA Space Apps challenge we picked was Commercializing Low Earth Orbit (LEO) — how to turn the post-ISS era into something more than a handful of private stations. Our angle: microgravity research today is locked behind multi-million-dollar entry tickets and dominated by agencies and large corporations. Even on Earth, RPM clinostats can’t fully neutralise gravity, distorting crystal growth and cell behaviour. Space BioLab makes the orbital lab a rentable resource — Where science reaches the stars.
What we designed
A 16U CubeSat carrying a modular payload focused on the two research lines with the strongest microgravity advantage, selected through a quantitative scoring framework (microgravity advantage 30%, CubeSat feasibility 25%, autonomy 20%, time-to-result 15%, market potential 10%):
- 3D organoids and spheroids — autonomous protocol, 60 days, optical characterisation. Microgravity yields ~99% cardiomyocyte purity vs. <90% in 1g, and ~2,000 differentially expressed genes in neural organoids cultured on the ISS.
- Protein crystallisation — micro-batch protocol, on-board imaging, and a deep-learning classifier that runs on the satellite to decide which images contain crystals worth keeping. JAXA’s GCF programme reports ~70% success rate in resolution improvement over 60+ missions.
Subsystems: modular microfluidics, integrated environmental control (37°C, pH, O₂/CO₂), automated optical imaging, internal reference centrifuge for dose-response curves (0.1–2g), and a multi-level FDIR (fault detection, isolation, recovery) layer with double biocontainment for BSL-1 organisms.
On-board crystal detection
A binary classifier (crystal / no-crystal) for the protein crystallisation payload. The premise: downlink bandwidth from LEO is the bottleneck, so the satellite has to triage its own images and send back only the ones worth a scientist’s time.
- Architecture: EfficientNetB3 pretrained on ImageNet, fine-tuned for binary classification. Inputs 300×300, Adam + binary cross-entropy, augmentation for rotation, flip, and brightness.
- Dataset: MARCO (Machine Recognition of Crystallization Outcomes), the standard benchmark in crystallography image classification. Balanced 50/50 across the two classes to avoid the bias of the original MARCO split.
- Validation results (epoch 6): accuracy 73.6%, precision and recall ~74%, F1 = 0.736.
- Deployment footprint: ~144 MB. Targets edge hardware (NVIDIA Jetson class) for on-board inference, no ground round-trip required.
Honest scoping: F1 around 0.74 is not flight-grade — for an operational mission you’d need a far larger and cleaner labelled set, hardware-in-the-loop validation, and a much more conservative decision threshold. The point of this work was to show that the AI component is technically reachable on a CubeSat budget, not that this exact model would fly.
Business model
A hybrid scientific SaaS, two tiers:
- Pro — $50–80k per slot for universities, startups, and education. Standard cartridge, processed telemetry, open-science publication path.
- Premium — $150–200k per slot for pharmaceuticals and biotech. Flight priority, X-band reserved bandwidth, longer rack time, real-time AI dashboards, NDA and partial exclusivity, dedicated payload engineers.
Target market: $1.2B (2023) → $5.4B (2032) in space-based biopharmaceuticals, CAGR 18.5%. Eli Lilly, Merck, and Novartis already run ISS programmes.
What I led
- End-to-end laboratory design — architecture, modular payloads, sensing and telemetry workflow, integration for microgravity use on the NanoAvionics M16P bus.
- Business plan from scratch — problem framing, value proposition, pricing tiers, market sizing, operating model, and milestones across a 4-year TRL 3→7 roadmap (~€1.05M total budget).
- Pitched the concept to judges and mentors during the hackathon weekend.
- Coordinated a six-person team across science, engineering, and business tracks over 48 hours.
Outcome
3rd place at the Valencia local event of the NASA Space Apps Challenge 2025 — a global hackathon organised by NASA together with ESA and 16 other space agencies, with 487k+ cumulative registrations across 190+ countries. Each local event runs its own competition; the top 2 from Valencia advanced to the global nomination round.
Team: Damián Vidal, Mar Minguez, Ignacio Pérez, Iago García, Mihail Stankevich, Carlos Mañes.