The AI-accelerated pipeline for personalized oncology
Bastard Software builds the compute infrastructure that translates raw tumor sequences into matched clinical trials, explainable treatment plans, and spatial GPU simulations of the tumor microenvironment.
Sequencing is cheap. Compute-driven interpretation is the bottleneck.
Hospitals generate terabytes of sequence data, but lack the HPC pipelines and AI reasoning to make it actionable. Actionable mutations are missed, and patients fail to match with life-saving trials.
Programy Lekowe — in the hands of Polish oncologists
The first OncoKernel tool is public, free and running at oncokernel.com: a reader for the Polish Ministry of Health drug programme criteria. No account, no tracking, no data collected — just the qualification criteria an oncologist needs, in a form they can actually read.
- Rebuilt from the Ministry's quarterly obwieszczenie, with a version stamp on every page
- Spot-checked against the source documents by a licensed physician, licence number published
- The reverse lookup nobody else offers — search by drug, find every programme it appears in
- 142
- drug programmes parsed from the quarterly announcement
- 41
- oncology programmes, classified from the Ministry's own index
- ~300
- reimbursed substances linked to official product information
- 0
- cookies, trackers or personal data collected
A unified ecosystem for clinical decision support
From instant trial matching for clinicians to deep GPU-accelerated tumor simulation.
Engineered for high-performance computing
GPU-native & AI-first
From accelerated bioinformatics pipelines to spatial SYCL-compute simulations, our stack is engineered for GPU acceleration — compiling to CUDA and integrating NVIDIA BioNeMo where they matter most.
On-premise & GDPR-compliant
Patient data never egresses to public clouds. Our containerized pipeline runs securely inside the hospital's network, ensuring full compliance by construction.
Grounded LLM Reasoning
Every clinical recommendation is backed by Semantic RAG, served locally on GPU, linking molecular profiles directly to cited biomedical literature and real-time trial registries.
Built by a clinician and an engineer
Founded by a physician and a GPU systems engineer, bridging the gap between clinical reality and high-performance computing.
Capital-efficient by construction
We wrap the gold-standard open-source bioinformatics the field has already validated rather than re-deriving solved problems — so engineering spend goes into the proprietary AI that is the actual moat.
A public front door that compounds
The free clinician tools are not marketing. They put us in front of the oncologists we need as design partners, and they prove out the reimbursement and trial data layer the on-premise product depends on.