Agent infrastructure for the biological layer of human life

AI is expanding into every vertical. The companies building that future will need structured, validated domain engines to reason against. GeneOps is that engine for human biology.

The Moment

Two exponential curves are crossing right now

Genome sequencing costs have collapsed faster than Moore's Law — from billions to under $200 today, with sub-$100 within reach.1 Simultaneously, AI agents have reached clinical-grade performance on medical reasoning benchmarks. This intersection creates a new category.

Sequencing cost collapse
Moore's Law
$2.7B → ~$200 110,000,000×
AI agent capability
Medical benchmark threshold
Pattern matching → Medical-grade reasoning2025–26
The Problem

50 million genomes. Almost none are operationalized.

50M+
People have raw genotype data from 23andMe and AncestryDNA.2 The data exists — accessible, exportable, and almost entirely unused.
~0%
Have received actionable, evidence-graded, gene-interaction-aware insights from their data. Most don't even know what's possible.
$100–450
Per genetic counseling session, uninsured.3 Weeks-long wait times. Not scalable. Not accessible.

Wellness platforms, fitness apps, supplement brands, and AI agents all want to personalize by genotype. None of them have the infrastructure to do it. The missing piece isn't consumer demand — it's the structured, validated intelligence engine that makes genomic personalization possible.

Demand

Accelerating on every axis

The tailwinds are structural. Falling costs, rising curiosity about personal biology, and a generational shift toward proactive self-optimization over reactive healthcare.

~25%
CAGR
Consumer genomics is one of the fastest-growing segments in health tech — projected to reach $18.8B by 2034, up from $2.5B today.4
$13B
DTC Genetic Testing by 2034
The DTC genetic testing market is projected to grow from $4.5B in 2025 to $13B by 2034.5 Nutrigenomics and health-risk testing are the fastest-growing segments.
32%
Personalized nutrition share
Personalized nutrition & wellness is the fastest-growing consumer health segment — now representing 32% of the broader personalized health market.6
The underlying dynamic is clear: sequencing costs are approaching zero, and an entire ecosystem of wellness platforms, fitness apps, supplement brands, and AI agents wants to personalize by genotype. They need an intelligence engine to do it. The infrastructure that organizes and operationalizes genomic data is where the value concentrates. That is GeneOps.
The Insight

AI needs a knowledge substrate to be useful in genomics

Every AI agent, wellness platform, and health product that wants to personalize by genotype needs the same thing: a structured, validated, continuously curated domain layer to build on. Without it, AI hallucinates and platforms guess. That substrate doesn't exist today. GeneOps builds it.

Without structured domain data

LLMs hallucinate about genomics

They confuse risk alleles with reference alleles. They miss gene-gene interactions. They cite retracted papers with confidence. No platform, agent, or product can build reliable genomic features on raw AI alone.

With GeneOps as substrate

Any agent, platform, or product becomes genomics-literate

Every recommendation grounded in published research. Evidence-graded. dbSNP-verified. Schema-validated with PubMed citations. Partners plug in and ship genomic personalization on day one — without building domain expertise.

The Shift

Software ate the world. Now agents are eating software.

Within 3–5 years, most people's primary interface with digital services will be an AI agent, not an app. But the domains where agents can be trusted are vanishingly few.

Domains where agents are trusted — a spectrum
Low stakes High stakes · high trust required Shopping Travel Finance Health Hardest
Low stakes High stakes · high trust Shopping Travel Finance Health Hardest
Low stakes → high stakesOpportunity
Trust Anchor

The highest-stakes, highest-trust, highest-value domain for agents

Universal
Every human has a genome. Every human has health questions. The addressable market is humanity. And your DNA doesn't change — a user acquired once has lifetime value.
Rigorous
You can't vibe your way through genomic interpretation. The wrong call on a clinically actionable variant has real consequences. This domain forces the kind of structured data layer that agents need everywhere — but few domains demand.
Ready
50M+ people already have their raw genotype data from 23andMe and AncestryDNA.2 No new hardware needed. The supply exists — the intelligence layer doesn't. Until now.

Most AI agent startups are building in low-stakes domains — shopping, scheduling, travel. We picked the hardest, highest-trust domain on purpose. Biology is where the rigor is non-negotiable, where the credibility barrier is highest, and where a working substrate is hardest to replicate. That's not a handicap. It's the moat.

Two Vectors

Agents for humans. Infrastructure for agents.

Two compounding vectors. One deploys AI agents into people's lives. The other builds the data infrastructure that all agents will need. Both compound with scale.

Vector 1 — Agents for humans

AI personalities that know your biology

Each vertical deploys an agent that knows the user's genome, remembers their context, and evolves with them. Allela is your wellness advisor. Darwin is your dating counselor. A GP-facing vertical puts a genomics agent in the doctor's office. Each vertical is a relationship, not a dashboard. We are getting extremely good at building agents that people trust with intimate, consequential decisions about their bodies.

Vector 2 — Infrastructure for agents

The data layer every health agent will need

When every health app has an AI agent, and every agent needs structured biological data to reason safely, we are the pipe they connect to. MCP-native from day one. The knowledge base is machine-readable by design — built for agents, not just for humans. The next wave of health technology will be built on top of infrastructure like this.

Agent-Native

Built for agents to wield and extend

Most genomics platforms were designed for a human to read a report. GeneOps was designed for an agent to reason against structured data. The dashboard, the API, and the MCP server all consume the same knowledge layer — the consumer product is a view, not the core.

// Any agent, any model, your genome

get_my_dashboard()
// → 9 categories, personalized profiles

get_my_actionables("nutrition")
// → Nutrition actions for YOUR genotype

get_my_genotype("rs1801133")
// → MTHFR C677T: CT heterozygous
// → evidence grade, actions, interactions

Full data sovereignty. Pair with a local model — Llama, Mistral, DeepSeek — and genetic data never leaves the user's machine.

Cross-reference anything. Hand your agent a blood panel, a prescription, a research paper. It reads them alongside the user's full genetic profile.

The most powerful way to organize your life. Power users plug their own agent into GeneOps and let it weave their genome into everything else it does for them — meal planning, training, supplement choices, conversations with their doctor. As agents take over more of daily life, this becomes the default interface to every service. GeneOps is just at the bleeding edge.

Alignment

Structurally aligned with the AI wave

We don't compete with AI labs — we build the domain layer they need. As models improve, as genetics research deepens, as consumer testing tech expands, the value of a deterministic, validated substrate compounds.

Complementary to AI platforms

AI companies build reasoning. We build the domain.

As AI expands into health, every major platform will need structured genomic intelligence to build on. This isn't something they can train into a general model — it requires deterministic, per-variant curation against published research. GeneOps becomes more valuable with every improvement in AI capability.

Synthesized, not scrapable

This data doesn't exist anywhere to crawl

Gene-gene interaction maps, evidence-graded variant recommendations, genotype-specific action plans — compiled from thousands of papers into structured schemas. It can't be scraped, prompted, or generated because it doesn't exist in this form until we build it.

Deterministic precision

Genomics demands exactness, not probability

You can't prompt-engineer your way to safe genomic recommendations. The wrong call on a clinically actionable variant has real consequences. This requires deterministic validation — not probabilistic generation. That's why AI companies will integrate this, not build it.

Landscape

Five incumbents, one shape — built for an era that ended.

Commercial genomics today is held by a handful of established players that made the correct architectural bets for the decade when sequencing was scarce, expensive, and clinical-only. Those bets are now structural commitments.

  • Dante Labs — owns its own EU sequencing lab; clinical + research bundle into hospitals and biopharma.
  • SelfDecode — API and white-label tiers, BYO genome supported; polygenic-risk centric, turnkey wellness template.
  • LifeNome — enterprise wellness; Fortune-500 logos, multi-year corporate contracts.
  • Sequencing.com — a marketplace; developers build on marketplace terms and don't own the end user.
  • GeneMetrics — wholesale kits and a clinician portal; independent clinicians, telehealth, med-spas.
Lab cost base

Pins the company to lab revenue

The 50M+ raw genotype files already in users' hands are a margin problem for a lab business — not an asset to build on.

Staff research team

Caps depth at human-team velocity

A salaried geneticist team caps how fast the knowledge base can deepen — and what it costs to deepen it at all.

Six-month sales cycle

Pins the customer to enterprise procurement

Hard to turn around in a quarter to serve a consumer-brand founder shipping in weeks, or an agent platform shipping in days.

Report-shaped product

Pins the architecture to human reading

Going agent-native on top of a presentation layer built for PDFs is a rebuild from the schema upwards — not a feature.

None of these are bad companies. They're competently run businesses that made sound bets for the era they were built in. The category isn't waiting for an incumbent to fix itself — it's waiting for a company built from scratch on the new economics. That's the opening.

The Route Through

We land in the open corner.
We become the infrastructure.

The opening move and the endgame are not the same company. GeneOps enters where the existing players don't compete — and matures into the layer they end up depending on.

Today — the opening market

Where the incumbents aren't looking

Online health personalities. AI-native consumer brands. The agent ecosystem. Partners who ship fast, own their brand, speak directly to audiences who already trust them, and have no in-house genetics team to retire. This corner sits outside the existing players' field of view — and it's where the next decade of consumer health is forming.

Tomorrow — the end game

The incumbents become customers

As GeneOps matures — more variants, deeper interactions, sharper agent integration, more verticals on the engine — the analysis and AI-tooling layer the incumbents run in-house stops being competitive with what GeneOps offers as a service. They keep their labs. The intelligence layer becomes ours. The platform stops being an alternative and starts being the substrate the industry runs on.

Asymmetry

50 million genotype files
are an opening, not a margin problem

The largest pool of human genomic data on Earth already exists — exported, downloadable, sitting in 50M+ inboxes from 23andMe and AncestryDNA. For a company with a lab to feed, that pool is a competitor. For a company without one, it's the runway.

If kit revenue carries the P&L

BYO files sit awkwardly inside the business

A lab-dependent business has to keep the kit funnel full to fund the lab. BYO users can be upsold to premium panels, but they remain a lower-margin segment — and the headline acquisition motion stays pointed at people who haven't been tested yet, not at the 50M+ who already have a file.

If software carries the P&L

Frictionless user acquisition

GeneOps offers white-label kits via partner labs where a partner wants them — and welcomes BYO files on equal footing because the platform's economics don't depend on the kit margin. A user uploads a file they already own and is in the product in minutes.

The Solution

The genomic intelligence engine that everything plugs into

A structured knowledge base of genomic research, paired with a deterministic matching engine and an agent-native API layer. Partners — wellness platforms, AI agents, labs, fitness apps — integrate once and ship genomic personalization immediately. We run our own consumer vertical as proof that the infrastructure delivers real value.

Knowledge

Structured, evidence-graded, per-variant research. Continuously expanding. The substrate that AI reasons against.

Match

Deterministic engine maps any genotype against the knowledge base. Exact and reproducible — same input, same output, every time.

Deliver

API, MCP server, white-label dashboard, or embedded widget. Partners choose the integration surface. We deliver the intelligence.

Act

Evidence-graded actions: supplements, nutrition, lifestyle, training, monitoring, avoidances — personalized to each user's genotype, not their demographic.

13,135
Personalized actions
2701+
Gene interactions mapped
41
Health categories covered
Expanding with every new study published
New research = new platform value
Inside the Engine

Every entry runs a gauntlet
before it ever sees a user

"Evidence-graded" is the marketing claim. This is the pipeline behind it. Each variant in the knowledge base passes through frontier-model research, multi-stage verification, and an independent audit cycle before it ships.

01

Research

Frontier-model agents under human curation. Per-variant deep research, schema-validated, version-controlled.

02

Allele identity

dbSNP cross-check. GRCh38 plus-strand orientation. Any ambiguity halts the entry.

03

Population sanity

Frequency cross-checks against gnomAD, 1000 Genomes, and TOPMED. Discrepancies block the entry.

04

Independent audit

A separate driver re-fetches every PubMed citation, re-checks orientation, re-grades the evidence.

05

Three clean runs

A batch ships only after three consecutive audit cycles produce zero findings. Anything less, it goes back.

This is what "deterministic" actually costs to build — and the reason a general-purpose model can't generate it on demand. The substrate becomes more valuable with every improvement in model capability, not less.

Product

What a partner delivers to their users on day one

This is our own consumer vertical — running live on the same infrastructure we offer to partners. Immediate value from automated actions. Infinite depth through AI. A partner integrating our API or white-label product ships this experience without building any of it.

→ Push · Actions from day one
Supplement

Use methylfolate — regular folic acid doesn't work well for you

Your body converts folic acid at roughly 65% efficiency. The active form bypasses the bottleneck entirely.

MTHFR C677T × COMT Val158Met · established
Training

You're built for endurance, not sprints

Your muscle fibers are shifted toward aerobic performance. Marathon over CrossFit.

ACTN3 + PPARA · moderate
Monitoring

Ask your doctor to check homocysteine yearly

Multiple gene variants raise your cardiovascular risk marker. Easy to monitor, easy to manage.

MTHFR + MTR + MTRR · moderate
← Pull · Ask anything, go deeper
You
Why shouldn't I take regular folic acid?
GeneOps
Your body is slower than average at converting folic acid into the form it actually uses — roughly 65% as efficient as normal. That means unconverted folic acid can build up and actually block the active form from working. PMID 22012854

This matters more for you specifically because you also have a slower-than-average system for clearing dopamine. These two things compound — less active folate feeds into an already backed-up process. PMID 26856749

What to do: Take methylfolate instead (400–800 µg/day). It's the pre-converted form, so your body can use it directly. Also worth avoiding fortified cereals and breads where practical — they use the synthetic form.
What about my B12?
How does this affect mood?
Show all related interactions
Vitamins & Detox 172
Nutrition & Metabolism 201
Brain & Mental Health 141
Medications & Drug Response 139
Fitness & Body 163
Immune & Gut 95
Heart & Inflammation 98
Hormones & Sleep 87
Skin & Eyes 80
Deployment

Turnkey deployment. Any vertical. Days.

Same engine, different face. Each vertical runs on the GeneOps intelligence infrastructure — knowledge base, matching engine, AI agent, payment, onboarding. Configuration, not development. These are all live today.

GeneOps Core Engine

Knowledge base · Matching engine · Agent layer · MCP server · Data platform

Prototypes — Live
AllelaB2C wellness companion · AI chat · health documents
GymGeneSports & fitness genetics · training-profile advisor
Genome XIPro-athlete genetics · explosiveness, endurance, injury risk · ongoing management + transfer evaluation
Darwin DatingDating service · offspring analysis · compatibility scoring
Pipeline — Near-term
Primary CareGP assistant
Longevity ClinicsPersonalized protocols
NutritionGenetic dietitian
PharmacyDrug interactions
GovernmentPopulation health

These aren't reskins — they're completely different products with different feature sets, different user flows, different business models. A B2C wellness companion and a B2B roster management tool share a genomic intelligence layer, but almost nothing else. That's the power of the platform: we can go from concept to deployed product in days of planning and design — at a speed that lets each partner ship a category-leading product without rebuilding the underlying genomics work.

Traction & Roadmap

Built and live. Where this goes.

Live now

Full intelligence engine operational — knowledge base, matching engine, API, MCP server

13,135 personalized actions, 2701+ gene interactions across 41 health categories

13 verticals live on the platform today

AI agent + MCP server production-ready for partner and external agent integration

Knowledge base browsable free at geneops.ai/research — public proof of domain depth

Business Model

How we capture value

Partner revenue is the core — every integration leverages the same knowledge engine at near-zero marginal cost.

B2B genomic intelligence API

$119B personalized health8

Labs, wellness platforms, fitness apps, and supplement brands embed our intelligence into their products. Per-query pricing or volume licensing. Instant genomic expertise from integration.

White-label deployments

$4.5B DTC testing5

Full branded experience running on our infrastructure. Partner owns the relationship, we power the intelligence. Revenue share or flat licensing.

Agent ecosystem

$2.7B AI in health9

As AI agents become the primary interface for health decisions, the platform that feeds them structured genomic data captures compounding value. MCP-native from day one.

Vertical subscriptions

50M+ potential users2

Each vertical prices independently for its market. Consumer wellness, sports, dating, localized markets — every vertical generates subscription revenue on the same engine.

Genotype-matched marketplace

$567B personalized health7

Supplements, vitamins, and products matched to each user's genotype. 10–25% commission. Available to partners as an embedded feature.

Phenotype data engine

Compounds with scale

Every user interaction generates genotype-phenotype correlations. We use it in-house to sharpen recommendations and deepen our gene-interaction maps — and as the dataset grows, to underwrite original research.

Flywheel
Partner integrations bring users → users generate phenotype data → data improves recommendations → better recommendations attract more partners → each partner deployment costs near zero to serve.
Operating Model

The platform was built the way the platform builds platforms

A conventional commercial genomics company carries a research team of geneticists, a lab, an enterprise sales cycle, and a roadmap dictated by the speed of those three things. GeneOps runs on a different operating model — and every layer you see in production was produced under it.

Knowledge base

Frontier-model agents, human curation

Per-variant deep research, multi-stage verification, an independent audit cycle. Production at a depth that a salaried research team would take years to match — and at a velocity a salaried team cannot match at all.

Platform & verticals

Agent-native engineering, every layer

Matching engine, REST API, MCP server, the consumer flagship at geneops.ai, and every partner vertical — built end-to-end with frontier-model agents in the loop. One technical founder producing the output of a team.

Go-to-market

The same model on the commercial side

Commercial colleagues working heavily with agents from day one. Tech builds bespoke tooling for the commercial workflow; commercial feeds tech the constraints worth solving next. The leverage compounds across functions.

Software-leveraged in place of headcount-leveraged — end to end. This is not a temporary stage to outgrow; it's the operating model the company is built on, and the source of the speed advantage that every partner integration inherits.

Team

A distribution engine and a software factory

CEO
Johan Stigels
Telecom industry veteran. Launched standalone consumer brands on shared infrastructure for a major Swedish telecom concern — the exact MVNO model now applied to GeneOps.
Built and scaled sales organizations — rapid deployment of go-to-market teams across consumer and enterprise segments.
Founded an investment company. Took a niche telecom operator public. IPO experience and public markets track record.
Active businesses in sports media. Deep network across media and sports industries.
CTO
Martin Eriksson
Software entrepreneur since 2008. Custom data management systems for government and institutional clients. Stockholm startup scene — founder, investor, board member.
Two decades leading development teams across enterprise software, media platforms, and data-intensive products. Launched multiple ventures along the way.
Since 2022: Deep focus on agentic AI. Decades of experience managing teams and complex projects channeled into orchestrating AI agents. Advising companies on AI adoption and building at the frontier.
Built the entire GeneOps platform — knowledge engine, vertical prototypes, MCP server, research pipelines — using AI agents.

Johan takes products to market at scale. He's done it in telecom, he's done it in media, he knows how to build distribution. Martin builds at AI-native speed. One person produced what would traditionally require a sizable team. Previous joint ventures between the two, including an ambitious media project with institutional investment. Deep mutual trust, proven working relationship.

Corporate & Compliance

A US corporate shell
on European data rails

GeneOps Inc. is the Delaware C-Corp parent. GeneOps AB is the Swedish operating subsidiary, where the team and the build are. Production infrastructure runs in the EU.

Delaware C-Corp
GeneOps Inc. — for US capital markets, acquisition optionality, and the corporate shape a future US acquirer expects to see on the cap table.
Swedish AB
GeneOps AB — the operating subsidiary, because Sweden is where the team and the building actually are.
EU residency
Production infrastructure runs in the EU — a hard requirement for selling into the European market, and a position the lab-heavy US incumbents can't easily mirror. The same architecture deploys to other regions when a partner's footprint requires it.
Compliance baseline
Encryption at rest and in transit, MFA, audit logging, a full DPA chain, right-to-erasure honoured end-to-end, and patient-data partnerships contracted under the applicable frameworks.

Structurally legible to US investors and acquirers. Operationally positioned to sell into the European market on day one. Two readings, one structure.

Sources

References

1
NHGRI Genome Sequencing Cost Data. Cost of sequencing a whole human genome, 2000–2026. genome.gov
2
Berkeley Genomics Institute, "How many human genomes have been sequenced?" — Private ancestry companies collectively claim ~50 million genotyped individuals. berkeleygenomics.org
3
Biology Insights, "How Much Does Genetic Counseling Cost?" — Out-of-pocket costs range from $100 to over $450 per session. biologyinsights.com
4
Precedence Research, "Consumer Genomics Market Size and Forecast 2025 to 2034" — $2.54B in 2025, projected $18.83B by 2034 at 24.95% CAGR. precedenceresearch.com
5
Global Market Insights, "Direct-to-Consumer Genetic Testing Market Report, 2025–2034" — $4.5B in 2025, projected $13B by 2034 at 12.4% CAGR. gminsights.com
6
Grand View Research, "Personalized Medicine Market Size" — Personalized nutrition & wellness segment held 32% revenue share in 2024. grandviewresearch.com
7
Grand View Research, "Personalized Medicine Market Size" — Global market estimated at $567.10B in 2024, projected to reach $1.2T by 2033. grandviewresearch.com
8
Precedence Research, "Precision Medicine Market Size" — $119B in 2025, projected $537B by 2035 at 16.26% CAGR. precedenceresearch.com
9
Towards Healthcare, "Precision Medicine Market Sizing" — AI in precision medicine sub-market projected at $2.74B in 2024, growing to $26.66B by 2034. towardshealthcare.com