AI & Machine Learning Company in Cambridge

Clinical and pharma scope is what separates a Cambridge AI budget from a generic one. A scoped AI proof of concept at Cambridge rates runs six to ten weeks and covers a data audit, a baseline model, an ICO data protection impact assessment, and a hosted demonstrator; a custom production ML model (clinical risk score, molecular property predictor, fraud engine, semiconductor design optimiser) adds MLOps, monitoring, a UK AI Bill aligned model card, and where applicable a DCB 0129 hazard log; a full production AI system brings RAG, multiple models, fine-tuning, and enterprise integration into AstraZeneca, GSK, or NHS estates. Cambridge rates sit above most UK regions because of the ARM and AstraZeneca talent premium and the depth of regulatory work required for pharma and clinical scope. We give fixed-fee proposals rather than open time-and-materials estimates, and we flag R&D Tax Credit and Patent Box eligibility at scoping.

For any AI system intended to inform clinical decisions or sit inside an NHS workflow, we run MHRA software-as-a-medical-device classification at discovery to determine whether the build is Class I, IIa, IIb, or III, then plan technical documentation against UK MDR 2002 (as amended) and where exported the EU MDR Annex II. We produce a DCB 0129 hazard log maintained by the supplier and a DCB 0160 clinical safety case maintained by the deploying organisation, both signed by a registered clinical safety officer. The NHS Digital Technology Assessment Criteria (DTAC) is treated as a parallel deliverable covering clinical safety, data protection, technical assurance, interoperability, and usability. We have shipped to Addenbrooke's and other Cambridge University Hospitals teams under exactly this pattern, and our documentation has passed NHS England digital assurance review without external rework.

Yes. We build target identification models, ADMET (absorption, distribution, metabolism, excretion, toxicity) property predictors, clinical trial enrichment and patient stratification models, real-world evidence pipelines on CPRD and SAIL data with appropriate IGARD and CAG section 251 approvals, and biomarker discovery systems on multi-omics datasets from the Wellcome Sanger Institute and Cancer Research UK Cambridge Institute. Our pharma deliveries run under GxP and GAMP 5 categorisation with validation deliverables (URS, FS, DS, IQ, OQ, PQ), 21 CFR Part 11 electronic records and signatures controls, ICH E9(R1) estimands documentation for any model feeding statistical analysis plans, and MHRA-ready model cards. We integrate with Schrödinger, RDKit, DeepChem, and AlphaFold 3 outputs where the chemistry team already uses them, rather than forcing a new toolchain on senior medicinal chemists.

The UK Artificial Intelligence (Regulation) Bill, introduced as a Private Member's Bill in the 2024 Parliamentary session and signalled by the Department for Science, Innovation and Technology (DSIT) in its February 2024 White Paper response, proposes a principles-based regime relying on existing regulators (ICO, MHRA, FCA, CMA, Ofcom, HSE) rather than a single horizontal authority. For Cambridge clients this means our delivery has to satisfy whichever sectoral regulator is in scope. Pharma triggers MHRA, fintech triggers FCA, employment scoring triggers EHRC, and any personal data triggers ICO under UK GDPR and the Data Protection Act 2018. Our discovery phase classifies the use case across these regulators, and we ship model cards, DPIAs, and bias audits structured to satisfy whichever combination applies. We also track the EU AI Act for clients exporting models across the Channel, since most Cambridge biotech and silicon clients have continental users.

When a project requires genuine novel research (new model architectures, geometric deep learning on molecular structures, foundation models for biology, rare-disease genomics, semiconductor design-space exploration beyond standard heuristics, quantum-classical hybrid pipelines) we scope collaborations with Cambridge Computer Laboratory faculty, the Cambridge Centre for AI in Medicine, the Department of Engineering, or specific spinouts in the Cambridge Science Park and Babraham Research Campus, rather than overselling in-house research capability. Our core team owns applied engineering, MLOps, regulatory documentation, and productionisation, which is where most Cambridge AI projects actually stall after a strong academic prototype. For standard production work (RAG, fine-tuning, classical ML, computer vision on established architectures, fraud and AML pipelines) no academic partner is needed and the cost stays inside a single Codazz statement of work.

We default to AWS eu-west-2 (London), AWS eu-west-1 (Ireland) for failover and disaster recovery, Azure UK South for clients standardised on Microsoft 365 and Azure OpenAI Service, and GCP europe-west2 (London) where the team already uses Vertex AI. For LLMs that must stay UK-resident (NHS PHI under the Common Law duty of confidentiality, FCA-regulated banking, MoD-adjacent classified work) we use Cohere's sovereign endpoints or self-host Llama 3, Mistral, and Mixtral on UK GPU instances. Cross-border processing is acceptable only where a documented Transfer Risk Assessment and Standard Contractual Clauses are in place under UK GDPR Article 46. We document residency in the model card and the data processing agreement so ICO auditors, MHRA inspectors, and NHS Information Governance teams have a clear answer rather than a vendor-attestation handwave.

A typical Cambridge biotech model (molecular property predictor, clinical trial enrichment classifier, real-world evidence cohort builder, biomarker discovery pipeline) takes sixteen to twenty-six weeks from kickoff to validated production, assuming clean curated data, ethics and IGARD approvals already in place, and GxP validation as part of scope. Weeks 1 to 4 cover data audit, regulatory classification, and baseline modelling. Weeks 5 to 14 cover modelling iteration, validation under GAMP 5, and where clinical scope applies DCB 0129 hazard log development. Weeks 15 to 22 cover MLOps, shadow deployment, 21 CFR Part 11 control verification, and internal quality assurance review. Week 23 onward is gradual rollout under change control. If labelling, ethics approval, or CPRD and Sanger data access is still pending, add eight to twelve weeks. We have shipped to this cadence inside AstraZeneca-grade GxP environments.

Most custom AI development in Cambridge qualifies for HMRC R&D Tax Relief. SMEs may claim under the merged RDEC scheme that came into effect for accounting periods beginning on or after 1 April 2024, replacing the prior SME and RDEC split, with an enhanced rate for R&D intensive SMEs. Eligible activity typically includes novel modelling, architecture experimentation, algorithmic and scientific uncertainty, systematic investigation, and failed-hypothesis work, rather than routine integration. We deliver time-tracked engineering logs, technical narratives, experiment histories, and uncertainty documentation aligned with the CT600L return and HMRC's revised guidance on software R&D. Where a Cambridge client patents resulting algorithms or molecular AI methods through the UK Intellectual Property Office, downstream profits may further qualify for the Patent Box 10 percent corporation tax rate. DSIT also runs targeted AI and life sciences grant programmes our clients often combine with R&D relief. Final eligibility sits with your R&D tax adviser and HMRC.