Personalized medicine was a research problem. We made it an engineering problem — and then solved the expensive part.
Navi8 screens medications against a person’s own proteins for single-digit dollars of compute, on a pipeline that is already running on our own hardware, inside an app that is already published, with two clinical pilots live. We are raising bridge capital to two specific, observable gates.
Three curves crossed in the last twenty-four months. What has not been solved is the part between them.
Structure-prediction models became good enough and open. Variant-versus-wild-type binding effects became tractable at scale. Sequencing became a consumer purchase. What nobody had built was the system that turns a person’s sequence and their real medication list into a bounded, cached, auditable set of GPU jobs that costs $9 instead of $1,625. That is Navi8.
Variant-selective architecture
Everyone else folds a proteome. We fold what changed — a 180× difference in compute per user, the difference between a loss on every subscription and a 97% modeled gross margin.
Cache-first data model
Reference structures, alignments and predictions are keyed on sequence hash and model version. The second user of a protein state is nearly free; renewals are almost pure margin.
Provenance by construction
Every candidate carries why it exists and every structure carries what produced it — what a regulator, a clinician and a payer each ask for first.
Same hardware. Same model. The only variable is how many structures the platform is forced to generate per user.
Anyone can download an open structure model. What is not commodity is a system that makes personalized screening cost $9 of compute against $300 of revenue — instead of $1,625 against the same $300.
Do not fold what does not change
Non-protein-changing variants never trigger structure generation — that removes the vast majority of genomic records from compute.
Deduplicate to unique sequences
Multiple VCF rows in one gene collapse to haplotype states. Repeat analyses reuse the cached structure by sequence hash.
Resolve only job-relevant proteins
Onboarding does not have to precompute a proteome before a medication analysis can start.
Reuse reference assets
Reference structures and alignments are shared across every subject and every compound.
Planning figures from the internal compute model at L = 256 and 80.7% parallel efficiency; a representative human protein length of ~430 residues implies an additional ~2.8× factor under an O(L²) assumption. Planning estimates, not production SLAs.
Three revenue lines, one engine.
The consumer app proves the science, the clinician dashboard proves the workflow, and the employer contract is where the money is.
Consumer & family app
$300 single · $900 family per year. Personalized screening, alternatives, interaction checking and household medication management. 80/20 single-to-family split in the model.
Clinician subscriptions & pilots
$12K → $120K per enterprise client. The dashboard proxies household management for a clinician panel. Two pilots running; a third closes the milestone gate.
Employer & carrier partnerships
Population fee + analytics. Self-insured employers carry ~10% YoY cost growth. The same app and pipeline, priced per covered life, white-labeled through PBMs.
Modeled figures from the Navi8 financial model, September 2026. Not a forecast of results.
Bridge capital to two observable gates.
When both are met, the priced round is a growth story rather than a science story.
A third site gives a comparative readout across independent clinician panels — what a payer needs before a population contract.
When paid users are limited by GPU capacity rather than demand, every added dollar converts directly into revenue.
Fleet expansion, employer launch and published clinical results, funded against demonstrated demand.
- Compute capacity — lift the user ceiling on our own GPU fleet.
- Production engineering — take the PHT-CSEIRA pipeline from working to shipped.
- Third clinical pilot site — the comparative read-out.
- Runway — to the priced round on demonstrated demand.
App published with sequence upload, medication tracking, screening mode selection and household management. Pipeline live on internal GPUs with a signed patient-haplotype fold. Provisional patent filed on modeling multi-drug interactions on a personalized proteome. Clean cap table: founders and advisors hold common; existing investors are on SAFEs at a $10M cap.
Four steps from this page to a signed SAFE.
Request the deck
Email invest@navi8.com for the investor deck and data room — financial model, technical walkthrough and pilot summaries.
Talk to the founders
A 30-minute call with our CEO and CTO, with a live walk through the app and the pipeline running on our cluster.
Sign the SAFE
Post-money SAFE at a $10M valuation cap, on standard terms. Minimum check discussed on the call. Documents are executed electronically.
Wire and close
Funds are wired to Navi8 Health, Inc. and countersigned documents are returned within two business days. Investor updates follow quarterly.
$400K remaining in the current SAFE.
Email invest@navi8.com or use the button to start the conversation.
This page is not an offer to sell or a solicitation of an offer to buy securities. Any offering is made only to accredited investors through definitive documents. Modeled figures are planning estimates and not forecasts of results.
Builders, clinicians and a compute architect.
A founding team that has shipped consumer software, run clinical practice and built supercomputers.