For CROs
Add the services biotech clients are asking you for.
Ophthalmic studies, advanced omics, and automated analysis and reporting run by agentic systems. Built on your infrastructure, then handed over so your team runs them.
What we add
Three services. Take one or all of them. Ophthalmology is our specialism; the omics and automation work apply to any study you run.
Ophthalmic services
Take on the ophthalmic studies biotech clients now send elsewhere. Retina is our hands-on research field; anterior segment and ocular safety are delivered with specialist partners.
- Disease modelsMacular degeneration, diabetic retinopathy, vein occlusion, geographic atrophy and inherited retinal disease; glaucoma and dry eye
- Imaging and functionOCT, fundus imaging, electroretinography and visual function endpoints
- Histology and molecularRetinal morphology, immunohistochemistry and target engagement
- Ocular safetyToxicology readouts, scoped with specialist partners
Advanced omics
Offer the readouts clients now ask for by name, with analysis designed around the biology of each study. Sequencing is a service you can buy; the analysis is the part most providers stop short of. We do that part, so you sell the readout without hiring a bioinformatics team and the sponsor gets an interpreted answer rather than a counts matrix.
- Bulk RNA-seqDifferential expression and pathway analysis
- Single-cell RNA-seqCell types, cell states, cell-specific drug response
- Spatial transcriptomicsExpression in its tissue context
- Multimodal integrationOmics linked to imaging and assay data
Automated analysis and reporting
Agentic systems take the study from raw files to a client-ready report, the same way on every study. They do the repetitive work; your scientists keep the sign-off.
- IngestYour instruments and formats, agreed at scoping
- AnalyseThe right statistics for the study design
- ReportPublication-grade figures, traceable methods
- Your QC gateYour scientists review and sign off each study
One study, from bench to report.
We have worked at the bench that makes the data, at the computer that analyses it, and with the sponsors who receive the report. The pipeline is built for all three.
Software versions, parameters and random seeds are pinned. Inputs, outputs and approvals are recorded for every run.
Improve what you deliver today
More consistent reports on the services you already sell, with your scientists back at the bench instead of the keyboard.
Expand what you can offer tomorrow
The pipeline is modular, so new readouts like single-cell and spatial plug into the same frame and become new service lines.
The framework
Built on the agentic AI mesh
We do not invent an architecture for each client. We use the agentic AI mesh, the pattern McKinsey’s QuantumBlack published for running agents at enterprise scale, and we bring the laboratory knowledge that tells each agent what good looks like in a study.
Why a tested pattern matters
Most laboratory automation fails the same way: a clever script, written by one person, that nobody else can run, audit or extend. The mesh exists because large organisations hit that wall at scale. It sets out how agents are registered, what each is permitted to touch, how they pass work between themselves, and how every action stays on the record.
Why the biology matters more
The architecture is the easy half. An agent that segments a retina, judges a dim ERG trace or picks the model for a repeated-measures design needs someone who has run those studies to define correct. That is the part we bring, and it is the part a generalist automation vendor cannot.
- What your lab producesOCT and fundus images, histology slides, ERG traces, plate reader and assay files, sequencing runs, LIMS records and protocols. Any format, any instrument.
- A mesh of specialist agentsIngest, quality control, image analysis, omics, statistics and drafting. Each does one job and hands on, so a new readout is one more agent rather than a rebuild.
- Shared study contextProtocol, arms, endpoints, acceptance criteria and past runs, available to every agent so the analysis matches the study that was actually run.
- Governed autonomyEvery agent registered, permissions scoped to what it needs, and every action logged. Nothing an auditor cannot follow.
- Your scientistReviews, decides exclusions and signs off. Judgement rather than data wrangling, and rejected work goes back to the mesh.
- One reportConsistent on every study, with every value traceable to the data it came from.
The agentic AI mesh is an architectural pattern published by McKinsey’s QuantumBlack in Seizing the agentic AI advantage, built on five principles: composability, distributed intelligence, layered decoupling, vendor neutrality and governed autonomy. We are not affiliated with McKinsey or QuantumBlack. We follow the published pattern and prefer open standards, so nothing we build ties you to one vendor.
How a build runs
Three tracks. Each one ends in something you can use, and you can stop after any of them.
Find the gap
- Sponsor work you decline today
- What your lab already has
- Which capability pays back first
Stand it up
- Models, assays, omics workflows
- Pipeline on your infrastructure
- Pilot on representative data, end to end
Make it yours
- SOPs and acceptance criteria
- Staff trained and signed off
- Qualification agreed with your QA
What you own
What you end up owning
Built around your systems. Facilities, LIMS, QMS and data systems stay in place. Any changes a build needs are agreed upfront.
Your sponsors stay yours. We don’t run a lab or sell studies. The work is under confidentiality, and any overlap with our biotech advisory clients is declared before we start.
Done before. We built ophthalmic disease models for a global pharmaceutical company bringing its testing in-house.
- New service linesOphthalmic studies · omics readouts · automated reports
- Capability we addModels · omics · pipeline · know-how
- Your labFacilities · staff · QMS · data infrastructure
What the handover includes
- An agreed workflow, and a specification of the inputs it supports
- Pilot on representative data or an agreed study
- Documented methods, software versions and configuration
- Defined QC checks and scientist approval points
- Acceptance results against agreed test cases
- SOPs, training and operating responsibilities
- Code and method ownership terms, agreed in writing
- A defined support period and change process
How the engagement works
Thirty minutes, no charge
No obligation either. Tell us what sponsors ask for that you can’t deliver yet, and we’ll tell you which capability is likely to pay back first.
Stop after any track
Each track is scoped and priced before it starts. No retainer and no minimum term. Equipment and consumables are bought by you, at cost, with nothing marked up.
A few builds at a time
Transfer takes real attention. We run a small number of builds in parallel, and we’ll say so if we can’t take yours on now.
Where to start
Send us a few details through the form. One free call is usually enough to tell whether a build makes commercial sense for your lab.
Source: H. Cheng et al., Ophthalmic drug discovery and development using artificial intelligence and digital health technologies, npj Digital Medicine, 2025.
References
The figures we quote, and where they come from. Market sizes are third-party projections, not a guide to what any one laboratory can charge.
- SNS Insider. Preclinical CRO Market, 2025–2033. Preclinical CRO spending of US$6.24B in 2025 rising to US$11.54B by 2033, with biopharma sponsors accounting for 63.6% of 2025 spending, outsourcing to preserve capital and limit fixed infrastructure.
- Credence Research. Ophthalmology Contract Research Organization Market. US$2.73B in 2024 rising to US$5.94B by 2032, a compound annual growth rate of 10.2%, with early phase development and laboratory services together making up 45% of service spending.
- Meticulous Research. Transcriptome Sequencing Market. Forecast to reach US$11.83B by 2035, with CROs and sequencing service providers the fastest-growing end-user segment at 8.9% a year, and bulk RNA sequencing the leading technology segment in 2025.
- Cheng H, et al. Ophthalmic drug discovery and development using artificial intelligence and digital health technologies. npj Digital Medicine 2025;8, 25 September 2025. Up to 60% less time in early-stage development where AI is integrated.
- Anaconda. State of Data Science survey. About 45% of working time spent collecting and preparing data rather than analysing it; other published estimates run higher.
- McKinsey & Company and QuantumBlack. Seizing the agentic AI advantage. The agentic AI mesh and its five design principles: composability, distributed intelligence, layered decoupling, vendor neutrality and governed autonomy.
- IQVIA Institute. Emerging Biopharma’s Contribution to Innovation (2022). Share of novel active substances originated outside large pharma, 2012 to 2021.
We are not affiliated with McKinsey or QuantumBlack. We follow the published pattern and prefer open standards, so nothing we build ties you to one vendor.
Managing Director & Founder
Dr. Adrian Cioanca
You deal with me from the first call to the handover. I have worked at the bench that makes the data, at the computer that analyses it, and with the sponsors who receive the report, which is why the pipeline is built for all three.
- PhD in retinal degeneration and bioinformatics; University Medal and the Frank Fenner Medal for the thesis
- Led a retinal research team; 28 peer-reviewed papers
- Co-founded a retinal therapeutics company: discovered the lead, secured the IP, raised the funding for safety and toxicology
- Built disease models and testing pipelines for small and large biotech, supporting client patents and US$3M in investment and grants
- Data and AI at a top global consulting firm, across government, energy and retail
Scope a capability build.
Tell us what your sponsors are asking for that you can’t yet deliver. We’ll tell you what it would take to say yes.
Get in touch → Book a free introductory call →adrian@tractumbio.com
The form takes a minute. A short note on the sponsor requests you are turning away is enough to start, and no confidential data is needed. The first call is thirty minutes, no charge and nothing locked in. We reply within two business days.