Clinical AI & diagnostics
Scope, build and validate AI models on clinical, imaging and laboratory data, with safety, ethics and human oversight considered from the first conversation.
Consultancy
Scientific and AI expertise, applied to your problem.
We work with hospitals, diagnostic labs, research groups and biotech teams to scope, build and validate AI and computational work.
What we can help with.
Engagements range from a focused advisory session to a full build-and-validate project. Every one starts with understanding your data and your constraints.
Computational drug discovery
Target identification, protein and ligand modelling, docking and molecular dynamics for research groups and early-stage biotech.
Health data & informatics
Turn EHRs, lab records and clinical text into structured data your team can analyse, report on and model, including biostatistics for clinical trials.
Bioinformatics & genomics
Sequence analysis, phylogenetics, primer design and genomic data workflows for labs and research projects.
Lab & infection control advisory
Practical guidance on laboratory workflows, quality processes and hospital infection control, drawn from active clinical practice.
Training for your team
Custom workshops for hospitals, colleges and companies, built from our EdTech bootcamps and taught by the same scientists.
Register for the 18-month diplomaOur first problem: anaemia.
We started where a single blood test most often falls short. Anaemia is common, chronic and has many causes, so reading it well means reading the patient’s whole clinical context.
1.92 billion
people live with anaemia, close to one in four worldwide.
Global Burden of Disease Study, 2021
- Chronic
- It develops slowly and lasts for years, so it is often caught late and managed for a long time.
- Multifactorial
- Nutrition, inflammation, genetics, bone-marrow disease and blood loss can each cause it, often several at once in the same patient.
- Context-dependent
- The same blood value can mean different things in different patients, so one number rarely explains the cause.
In preparation
Context-Aware Haematological Intelligence I: Utility and Discordance of Iron Biomarkers in Anaemia
US adult population data
What we found
Ferritin, the iron test ordered most often, stops tracking anaemia once inflammation is present, and adding it can make models worse. Other iron markers held up far better across patient groups, and in some groups the markers often disagreed with each other.
In preparation
Context-Aware Haematological Intelligence II: Disease-Mediated Remodelling of Anaemia in Chronic Kidney Disease
US adult population data, grouped by kidney function
What we found
In chronic kidney disease, anaemia runs on different biology: kidney-function markers overtake iron markers, and models lose accuracy as kidney function falls. The stakes are high, as adults with both CKD and anaemia had markedly higher mortality than those with neither.
In preparation
Context-Aware Haematological Intelligence III: Differentiating Thalassemia from Iron-Deficiency Anaemia
Several datasets from around the world
What we found
Routine blood counts can flag beta-thalassaemia with high accuracy, even with simple models, using values labs report but rarely read. Alpha-thalassaemia carriers stayed hidden: no routine blood count parameter reliably separated them.
All three manuscripts are in preparation. The cohorts behind them come from outside India, so our next step is validation on Indian hospital data. If you hold haematology records, or want a pre-publication briefing, talk to us.
Discuss the researchStrategic partnership with ClinEra
We are ClinEra’s biostatistics partner for a pilot clinical trial, and built its analysis pipeline end to end, from data checks to the efficacy readout. The outcome data were skewed and far from normal, the situation where standard tests give misleading answers. Our pipeline used methods that don’t rely on normality, and established the treatment’s efficacy on the data as it really was.
- Scope
- Biostatistics pipeline for a pilot clinical trial
- Challenge
- Non-normal, heavily skewed outcome distributions
- Result
- Efficacy established with methods that hold without assuming normality
How an engagement runs.
A clear scope up front, regular check-ins along the way, and a handover your team can build on.
-
Step 1: Conversation
Tell us the problem, the data you have and what a good outcome looks like.
-
Step 2: Scoped proposal
We come back with an approach, deliverables, timeline and cost.
-
Step 3: Build and validate
We do the work with regular check-ins and test it against your real conditions.
-
Step 4: Handover
You receive the outputs and documentation, plus training for your team where it helps.
Questions about working with us.
Short answers to what hospitals, labs and research teams usually ask before an enquiry.
Who do you work with?
Hospitals and diagnostic labs, universities and research groups, biotech and pharma teams, and health-tech companies.
How much does a project cost?
Every engagement is scoped individually. After a first conversation, we send a proposal that sets out the approach, deliverables, timeline and cost. Work ranges from a focused advisory session to a full build-and-validate project.
How do we get started?
Send a project enquiry or email queries@omnihemei.com. A member of the team replies by email to set up a first conversation about the problem, the data you have and what a good outcome looks like.
Do you do clinical trial biostatistics?
Yes. We are ClinEra’s biostatistics partner and built the analysis pipeline for its pilot clinical trial, from data checks to the efficacy readout, using methods that stay reliable when outcome data are skewed or non-normal. Read about the ClinEra work.
What research has OmniHemEi done?
Our first research series, Context-Aware Haematological Intelligence, uses machine learning on routine blood tests and patient context to read anaemia more accurately. Three manuscripts are in preparation: on iron biomarkers, on anaemia in chronic kidney disease, and on telling thalassaemia apart from iron-deficiency anaemia. See the findings.
How do you handle safety in clinical AI projects?
Safety, ethics and human oversight are considered from the first conversation. Scoping covers software as a medical device (SaMD) considerations, and models are validated against your real conditions before handover.
What do we receive at the end of a project?
The outputs and documentation, plus training for your team where it helps. We also run custom workshops for hospitals, colleges and companies, built from our EdTech bootcamps.
Start an enquiry
Tell us about your project.
Share a few details and a member of the team will reply by email to set up a first conversation.
- Hospitals and diagnostic labs
- Universities and research groups
- Biotech and pharma teams
- Health-tech companies
Prefer email? queries@omnihemei.com
Project enquiry
Enquiry sent
Thanks. A member of the team will reply by email to arrange a first conversation.