One pipeline, growing set of tools
Each tool is built to stand on its own, and together they form a single continuous path — from the first search for a nearby trial, to a full molecular interpretation of the tumor, to simulating and designing the therapy itself. More tools join this pipeline over time.
Find the closest matching clinical trial in seconds
A public-data panel built for oncologists. Enter an indication, a few biomarkers, and a location — get back the nearest recruiting clinical trials and the drug reimbursement programmes a patient would qualify for. Europe first: EU trial registries (CTIS, EUCTR) unioned with ClinicalTrials.gov, plus national drug reimbursement schemes.
- Ranked by clinical fit and distance to the patient
- Includes national drug reimbursement programmes, not just trials
- Public registries only — no patient sequence ever touched
Phase II — Adavosertib in TP53-mutant sarcoma
B.72 drug reimbursement pathway — bone sarcomas
OncoKernel: from raw tumor sequence to a citation-grounded treatment plan
OncoKernel runs entirely inside the hospital's own network. It takes a tumor sample through four decoupled blocks — each replaceable, each testable on its own — and ends with a physician-facing dashboard and a grounded chat copilot.
Raw reads to a characterised genome
Somatic variant calling, tumor purity/ploidy/TMB/MSI, and HLA typing — wrapping clinically-validated, open-source tools rather than re-deriving solved bioinformatics.
Annotate and score every variant
Known variants are resolved against a live graph of genes, drugs, and indications; variants of uncertain significance are scored with a genomic foundation model.
Match genotype to therapy
A multi-agent retrieval system matches the patient's molecular profile against guidelines, literature, and live trial registries — every claim cites its source.
Present it to the physician
A scannable dashboard with a grounded chat copilot and one-click PDF export — informational decision support, the oncologist decides.
GDPR by construction
Genomic data never leaves the hospital network — the same architectural choice that unlocks the sale unlocks the compliance story.
Genomic foundation models
Zero-shot variant-effect scoring via Evo 2, GPU-accelerated, for mutations no database has classified yet — with a plain-language rationale attached.
Live knowledge graph
Genes, variants, drugs, pathways, and indications resolved against a continuously updated graph, not a static spreadsheet.
Grounded clinical copilot
Served locally for low-latency inference — no cloud round-trip. Every answer cites the exact source paragraph; informational decision support, the oncologist always decides.
A GPU-native Digital Twin of the tumor microenvironment
Our Agent-Based Modeling (ABM) engine is built compute-first with SYCL — a portable compute standard across GPU vendors. On NVIDIA hardware it compiles to native CUDA for maximum performance. It dynamically simulates cancer cells, immune infiltration (TILs), and drug diffusion as continuous fields, executing entirely on GPU compute shaders — letting us test AI-generated therapies in-silico before they ever enter the wet lab.
SYCL Compute, GPU-Native
Massively parallelized spatial hashing and collision detection on GPU compute shaders — eliminating CPU bottlenecks typical of legacy simulators.
Biologically grounded agents
Simulating hypoxia, angiogenic signalling, and CAR-T cell penetration based on the genomic profile extracted by OncoKernel.
In-silico therapy screening
Providing a deterministic, physics-based environment to evaluate the kinetic efficacy of immunotherapies designed by generative AI.
AI + Wet Lab: Bringing in-silico designs to physical reality
While our software stack models the biology, our ultimate roadmap involves physical validation. In the future, the neoantigens identified by OncoKernel and folded by BioNeMo will be synthesized into mRNA vaccines and validated in patient-derived organoids and tumor-on-a-chip microfluidics in our partner wet labs.
AI-designed neoantigen vaccines
mRNA transcripts designed against a patient's tumor-specific mutations and HLA type, validated in-silico before synthesis.
CAR-T therapy design
Engineering chimeric antigen receptor constructs targeting the tumor-specific antigens the platform surfaces for that patient.
Wet-lab validation
Patient-derived organoids and tumor-on-a-chip models give a physical check on every AI prediction before it reaches a patient.