AminoClaw
The AI Co-Scientist for Protein Design

From a single message to a validated protein binder —
in hours, not months. No coding. No specialist required.

Seed Round 2026
Raising £2M
£8M Pre-Money
London, UK
Confidential
hello@agenticbio.ai  ·  aminoclaw.acetyl.online
The Opportunity

Proteins are the molecules of life.
And the industry that designs them is broken.

$2.6B
average cost to bring
one drug to market
90%
of drug candidates
fail in clinical trials
12+
years from discovery
to market approval
$700B
global biologics
market by 2030
The Industry Signal
  • 78 novel drugs approved globally in 2024 — accelerating year on year (IQVIA 2025)
  • 65–75 new medicines expected annually through 2029
  • Oncology, antibody therapies, enzyme biologics: fastest-growing categories
  • $181B in new brand spending growth over the next 5 years
The AI Convergence
  • AlphaFold3, Boltz-2, RFdiffusion all reached production maturity in 2024–25
  • Agentic AI market: $200B+ by 2034, CAGR ~44%
  • Gartner: 40% of enterprise apps will embed AI agents by end of 2026
  • VC funding in agentic AI: $9.7B+ since 2023 — capital is flowing
The tools exist. The models are mature. The missing piece is an agent that orchestrates all of it.
Who We Serve

One platform. Three enormous markets.

Any organisation that works with proteins — from a PhD lab to a global pharma company.

Academic Labs
100,000+
protein research labs worldwide
The pain

Every biology PhD knows proteins — but fewer than 1% can run a frontier AI design pipeline. AlphaFold3, RFdiffusion, ProteinMPNN all require specialist coding skills most wet-lab biologists simply don't have.

Our value
  • Chat-based interface — any biologist can use it
  • Publication-ready analysis in hours, not weeks
  • Academic pricing tier: accessible on grant budgets
  • Direct pipeline into high-impact publications & IP
Revenue model
£2K–10K / lab / year · SaaS
In-Silico CRO Platform
$400B+
global CRO & protein services market
The pain

Traditional CRO: £50K–500K per campaign, 3–6 months per iteration. Pharma R&D teams are drowning in bottlenecks. The same problem exists in enzyme biosynthesis, cosmetic peptide design, and industrial protein engineering.

Industries we serve
  • Biopharmaceuticals — antibody & binder optimisation
  • Biosynthesis — enzyme engineering for chemicals, biofuels, food
  • Cosmetics & aesthetics — collagen variants, peptide actives, growth factors
  • Industrial enzymes — plastic degradation, textile processing
Revenue model
£10K–50K / month enterprise · 80%+ margin
Academic → Industry Bridge
$130M
vs. $1.1B — AI super agent preclinical cost reduction*
The pain

Academia discovers frontier protein targets. Industry needs validated drug candidates. The gap — translating a discovered target into hit molecules — traditionally costs £2–5M and takes 18+ months of CRO work.

Our value
  • Academic lab publishes target → AminoClaw generates 100s of candidate binders
  • Selectivity, affinity & ADMET ranked in-silico — top candidates ready for wet lab
  • Compress 18-month CRO pipeline to weeks at a fraction of the cost
  • Revenue on milestone: target → hit → lead
Revenue model
Success fee + milestone payments on pipeline progression
* Source: 途深智合 AI Drug Discovery Market Analysis, 2026
The Problem

The bottleneck is not biology.
It's the expert who can't scale.

<1%
life scientists can independently run
an AI protein design pipeline
3–5 months
per manual protein design campaign
using traditional methods
£80–150K
annual cost of one bioinformatics
specialist — per person, per year
Why it's broken today
  • AlphaFold3, RFdiffusion, ProteinMPNN, AutoDock — each brilliant at one thing, each requiring specialist setup
  • 5+ separate tools with incompatible data formats and compute environments
  • Manual handoffs between tools introduce errors, delays, and reproducibility failures
  • A single design iteration that AminoClaw does in hours takes a team weeks
  • Result: only well-funded labs with dedicated bioinformaticians can access frontier AI
The opportunity cost
  • Biotech startups spend £500K+ before they even have a lead candidate — most of that is wasted iteration
  • Academic labs sit on publishable discoveries because they lack the compute expertise to validate them
  • Pharma R&D cycles take 18+ months for what should be 3-month campaigns
  • The world's frontier AI models exist — but 99% of biologists cannot access them
Assistants answer questions. Agents own the entire workflow. AminoClaw is the first protein design agent.
Product — AminoClaw

One conversation. The entire protein
design pipeline — automated end-to-end.

01
📚
Literature
PubMed, bioRxiv, Semantic Scholar — prior art synthesis & epitope mapping
02
🗄️
Data Retrieval
PDB, UniProt, AlphaFold DB — structures fetched, cleaned, aligned automatically
03
🔬
Analysis
DMS parsing, fitness scoring, conservation across 150+ orthologs, hotspot detection
04
🗺️
Mapping
Interface prediction, selectivity handles, druggability scoring, binding site ID
05
⚗️
Design & Validate
RFdiffusion→ProteinMPNN→AF3/TEDstruct cross-validation, 3 automated refinement rounds
06
📊
Report
Publication-ready report + interactive 3D viewer delivered to Slack / email / API
211
AminoClaw skills
4
proprietary models
20+
AI models orchestrated
Zero code required. A biologist describes their target in natural language — the agent plans, executes, validates, and iterates across the full design cycle.
Hours, not months
<$100 per pipeline run
Proven Results

Not demos. Real designs.
Quantified, reproducible outcomes.

Binder Design
RBX1 E3 Ubiquitin Ligase
Cancer target — novel binder designed from scratch
+116%
ipSAE score improvement · 0.30 → 0.65
  • 88aa mini-binder, active-site targeting
  • 3 automated rounds of AF3 cross-validation
  • Completed in hours vs. 3–5 month manual benchmark
  • Cost: <$50 in cloud compute
Enzyme Optimisation
LCC Plastic-Degrading Enzyme
Industrial enzyme — maximise thermostability & activity
9,426
variants screened automatically
  • Top hit N215H: +1.84× activity vs wildtype
  • 25 elite hotspot positions identified
  • Zero manual curation — fully autonomous
  • Applicable to any enzyme engineering project
Selectivity Engineering
SlUGT91R1 Plant Enzyme
Substrate selectivity — 10⁵× gain achieved
10⁵×
selectivity improvement · 6 ranked mutants delivered
  • RF3 + AutoDock Vina cross-validated pipeline
  • Fully automated design → score → rank workflow
  • Generalises to any protein substrate pair
  • Complete 3D structural analysis included in report
The same pipeline generalises to any protein target — drug biologics, industrial enzymes, cosmetic peptides, biosynthetic pathways.
Our Technology — Proprietary IP

We don't just call public APIs.
We have our own models.

Four self-developed models that make our agent smarter than any competitor using off-the-shelf tools.

TEDlm
Protein Language Model
Foundation model capturing structural and functional properties at sequence level. Powers the agent's deep protein understanding — not just pattern matching, but physical insight. Developed by Tiejun Wei.
VariPred
Variant Effect Predictor
Predicts how mutations affect protein function, stability, and binding affinity. Powers hotspot detection and mutation scoring across our AminoClaw skills. Co-authored by Weining Lin and Dr. Jude Wells (Imperial).
TEDstruct
Structure Prediction Model
Complementary to AlphaFold3 — particularly strong on designed and non-natural sequences that AF3 struggles with. Enables accurate validation of in-silico designed binders before wet-lab testing.
FluxAb
Antibody Design Model
Specialised for antibody sequence optimisation — CDR region design, affinity maturation, immunogenicity reduction. Integrates with ProteinMPNN for full antibody design pipeline with wet-lab validated outputs.
ProFam
Protein Family Language Model
Family-aware protein representation model trained on clustered sequence space. Enables evolutionary context-aware design, function annotation, and cross-family generalisation beyond single-sequence models.
Market Size

We are entering multiple large markets
at the exact right moment.

TAM
$200B+
Agentic AI market by 2034
CAGR ~44%
SAM
$68B
Bioinformatics + AI drug discovery
market by 2035 · CAGR 13–15%
SOM (near-term)
$10B+
AI protein design services by 2030
Our primary beachhead
Bottom-up: our reachable market now
Academic labs (100K+)
£200M
Biotech (5–50 person)
£600M
Pharma R&D teams
£2.4B
Industrial / cosmetics
£800M
Total reachable: ~£4B within protein design services (conservative, 2026 figure)
Why winner-takes-most dynamics
  • Whoever owns the agent layer owns the workflow — users don't switch once embedded
  • Each new skill and case study compounds the data advantage
  • First-mover in biology-native agentic AI = dominant default for protein labs
  • Comparable: GitHub Copilot now has 77% of AI coding assistant market after 3 years
  • Our FRS advisor network gives warm enterprise access that cannot be bought
Business Model

Three revenue streams.
Recurring, high-margin, naturally expanding.

Academic SaaS
£2K–10K
per lab / year
  • Access to AminoClaw agent & core AminoClaw skills
  • Academic pricing — accessible on grant budgets
  • Publication-ready outputs + citation-ready methodology
  • Natural upgrade path to enterprise tier on spinout
Target: 500 labs by end 2027
= £2.5M ARR at £5K avg
PRIMARY
Enterprise CRO
£10–50K
per month · per workflow
  • Custom agent workflows for specific pipelines
  • Private cloud / on-prem deployment
  • SOC-2 & HIPAA compliance, full audit trails
  • Dedicated SLA, white-glove onboarding
Target: 10 enterprise clients by Q3 2026
= £500K–1M MRR at scale
Outcomes / Translation
Success fee
+ milestone payments
  • Revenue share on agent-delivered drug candidates
  • Milestone: target → hit → lead → IND
  • Aligns incentives with pharma & biotech timelines
  • High-value per deal: £200K–2M per successful transition
  • Hit-to-lead: traditionally 12–15 monthsAminoClaw compresses to 4 months
Even 1 successful transition / quarter
= significant recurring high-margin revenue
Financial Projection

3-Year ARR Projection

Conservative targets based on current pipeline and GTM plan.

2026
£200K
→
2027
£1.2M
→
2028
£4.5M
£4.5M £3M £1.5M £0 2026 2027 2028
Gross margin 80%+ Net Revenue Retention 140%+ ● Path to breakeven ~2027
Traction & Partners

Real customers. Strategic partners.
Early proof of demand.

🧬
YOHOLIFE
First Paying Customer

London biotech designing & optimising enzyme specificity. Using AminoClaw to accelerate their PURELIX synthetic biology pipeline — enzyme screening and biosynthetic pathway construction.

🤖
MiniMax
LLM API Partner

China's leading multimodal AI (HKEX-listed, $4B+ valuation). Direct CEO relationship. Provides frontier reasoning API at 10× lower cost than US alternatives — our key to capital-efficient scaling.

☁️
Alibaba Cloud
Cloud Infrastructure

Global cloud infrastructure partner delivering discounted GPU compute and storage, optimised for large-scale AI and bioinformatics pipeline execution at scale.

1
paying customer
3
strategic partners
211
production skills
3
validated case studies
4
proprietary models
Roadmap to Series A
Q1 2026 ✓
First revenue
YOHOLIFE signed
Q2 2026
5 enterprise
design partners
Q3 2026
£500K ARR
SOC-2 certified
Q4 2026
Series A ready
£5M–8M target
Team

The bridge between frontier science
and production-grade AI engineering.

Tiejun Wei
Tiejun Wei
CEO & Co-founder
Computational Physics → AI Drug Discovery
  • Peking University B.Sc. · UCL M.Sc. · QMUL Ph.D. in Computational Science
  • UCL Research Fellow — quantum chemistry, biophysics, bioinformatics
  • Developer of TEDlm (protein language model) & TEDstruct
  • Extensive de novo protein design and molecular simulation experience
  • Bridges deep computational physics with AI-driven drug discovery
Weining Lin
Weining Lin
CSO & Co-founder
AI for Protein Design
UCL B.Sc. · MRes Neuroscience · Ph.D. AI Computational Biology
  • UCL AI Centre Research Fellow — AI for protein science
  • Co-developer of VariPred & FluxAb · Author of ProFam & GOBeacon
  • Expert in antibody & enzyme optimisation, directed evolution & de novo protein design
  • Bridges wet-lab protein biology with large-scale AI modelling
Scientific Advisory Board
David Jones
Prof. David T. Jones FRS
UCL · AlphaFold team · PSIPRED creator · 67K+ citations
Christine Orengo
Prof. Christine Orengo FRS
UCL · CATH database creator · 49K+ citations · EMBO Member
Jude Wells
Dr. Jude Wells Ph.D.
Imperial College · DeepMind hackathon winner · ProFam co-author
The Ask

Raising £2M Seed at £8M pre-money
to own the protein design agent layer.

Use of Funds — £2M
Platform Engineering & Skills 40% · £800K
Agent platform, 300+ skills, multi-agent orchestration R&D
Wet Lab Integration & Validation 25% · £500K
Accelerate agent pipeline with wet-lab feedback loop; improve model accuracy
Go-to-Market — 10 Enterprise Partners 20% · £400K
Sales, BD, pharma/biotech design partner acquisition
Compute Infrastructure Deployment 10% · £200K
GPU cluster, model serving infrastructure, SOC-2 certification
Operations & Legal 5% · £100K
£8M
pre-money valuation
20%
equity offered
18 mo
runway to Series A
Why we need this capital
Platform — scale from 211 to 500+ skills; build the multi-agent orchestration layer that no competitor has yet
Wet lab loop — close the dry→wet feedback cycle; every result improves our models and builds proprietary training data that cannot be replicated
Enterprise sales — our 2 FRS advisors open the door; we need a team to walk through it and close pharma design partner agreements
Compute — deploy dedicated GPU infrastructure for enterprise private deployment, a hard requirement for regulated pharma clients
12-Month Milestones
Month 0 — First paying customer (done)
YOHOLIFE signed, 3 case studies live on aminoclaw.acetyl.online
Month 3 — 5 enterprise design partners
Signed biotech/pharma clients at £10K–30K/month via advisor network
Month 6 — SOC-2 certified + wet-lab loop live
Enterprise procurement unlock; first dry→wet→model feedback cycle complete
Month 12 — Series A ready
£1M+ ARR · NRR > 140% · 10+ enterprise clients · £5M–8M raise
Remember Three Things
1
The only end-to-end protein AI agent that actually ships results.
211 production skills. Sub-$50 per campaign. Hours not months. A paying customer already. While competitors are still stitching tools together, we're running.
2
Five proprietary models. A data flywheel no one can buy their way into.
TEDlm · VariPred · TEDstruct · FluxAb · ProFam — built in-house, not rented. Every wet-lab result feeds back into training data our competitors will never have.
3
The agent layer in biology is winner-takes-most. We're first.
100,000+ labs. $400B+ market. Two FRS fellows who shaped the field. The window to own this category is open right now — and we're already through the door.
AminoClaw
London, UK · hello@agenticbio.ai · aminoclaw.acetyl.online
6 / 14