B2B research-compute systems studio
AI and data research systems, built and tested from hardware to handover.
GenomeRx builds the physical compute and the bioinformatics, data analysis, AI and LLM workflows that run on it, on macOS, Windows or Linux. One studio is accountable for all of it: architecture and code, compute integration, reproducible analysis, LLM evaluation, acceptance evidence and a documented handover, then scoped support if wanted. Research use only.
From research question to supported handover.
Five stages in one sequence. Each stage names its action, who does it and where, and the output it hands on. Select a stage for its purpose, its deliverable and the client decision that closes it.
- Stage 01Define Frame research need Researcher + delivery team Output: Scope brief + requirements
- Stage 02Design Define workflow and target platform Integrated compute Output: Architecture + acceptance plan
- Stage 03Build Configure platform and version workflow Integrated compute Output: Build specification
- Stage 04Verify Run client-agreed acceptance test Integrated compute Output: Test record + acceptance evidence
- Stage 05Handover Transfer to operations Operators + researcher Output: Runbook + training + support plan
Stage 01, Define
- Purpose
- Fix what the system must do: the research question, the data and its size, where the system has to run, and the acceptance criteria.
- Deliverable
- A scope brief with the requirements and the acceptance criteria, agreed in writing.
- Client decision
- Approve the scope, or change it before design starts.
On this page Contact
Stage 02, Design
- Purpose
- Choose the workflow design, the compute and the software stack that meet the scope, and write down how acceptance will be tested.
- Deliverable
- An architecture document and an acceptance plan: test cases, inputs, expected outputs, runtime and resource limits.
- Client decision
- Approve the architecture and the acceptance plan, including any hardware to be quoted.
On this page Platform
Stage 03, Build
- Purpose
- Write the workflow as versioned code and configure the workstation, appliance or server it runs on.
- Deliverable
- The code repository, the environment and build specification, and the configured compute.
- Client decision
- Review the build at the agreed milestones. A change of scope returns to Define as new scope.
On this page Build
Stage 04, Verify
- Purpose
- Run the workflow end to end on the built hardware, in the environment where it will live, against the acceptance plan.
- Deliverable
- A test record: outputs, runtime and resource use compared with the agreed criteria, and a rerun that reproduces the result.
- Client decision
- Accept the system, or return failed tests to Build.
On this page Test
Stage 05, Handover
- Purpose
- Hand the system to the people who will run it, with the documents they need to operate and change it.
- Deliverable
- Runbooks, a walkthrough for the operators, the test record, and support under a monthly retainer if wanted.
- Client decision
- Confirm the handover and decide whether support continues.
On this page Handover and support
Supported operation
The end state a delivery works toward: the accepted system runs under its own operators, with the test record, the runbook and the training behind it, and support under a retainer if wanted. Changes inside the agreed scope follow the improvement loop, one approved change at a time.
Connect
Share a workflow, platform constraint or delivery need.
The workflow and the machine, built as one system.
Both halves are scoped together, so the code and the hardware answer the same requirement.
Workflow and code construction
Bioinformatics, data analysis and AI pipelines are written as versioned code: orchestration, analysis, model evaluation and reporting. Runs are measured on the real workload.
Compute and hardware integration
Workstations, appliances and servers are specified and assembled for that workload: accelerators, storage, networking and the software stack. Hardware is supplied only inside a scoped B2B system, never resold on its own.
Platform scope
Built for the platforms you already run.
Environments, languages, model families and sourcing routes a scoped GenomeRx system can use, how LLMs are integrated and tested alongside the code and compute, and how a workflow changes after handover. Names are examples, not partnerships or endorsements. The written scope sets the final selection.
- Environments
- macOS, Windows and Linux. Hardware can be client-owned, sourced for the project, or installed where the system will run.
- Development
- Python for pipelines, services and AI workflows. R for computational life sciences and research analysis. SAS for biostatistical analysis.
- Hardware procurement
- Sourced through the OEMs, distributors, enterprise suppliers and retailers that fit the build, for example Apple, Dell, Microsoft, Best Buy and Micro Center where suitable. These are examples, not affiliations. Hardware is supplied only inside a scoped B2B system, never as standalone retail resale.
- AI models
- Client-approved frontier API model families such as Claude, GPT and Gemini, and selected open-weight families such as Llama, Mistral, Qwen, DeepSeek, Gemma, Phi and Nemotron, integrated and evaluated on the client's task. Availability, licensing, client access, runtime fit and evaluation results decide the final selection.
- LLM integration and evaluation
- Client-approved models are integrated into research and data workflows. Each integration has a repeatable evaluation harness: representative data and tasks, versioned prompts and configurations, regression checks in continuous integration, human review checkpoints and saved evaluation artifacts. Client requirements, access, licensing, evaluation results and review decide the final model and workflow. After handover the workflow changes only through the loop below, one approved change at a time. Nothing in it learns or changes on its own.
Governed improvement loop After handover
- Version
- Evaluate
- Review
- Release
- Monitor
- Next approved change
Research and analysis areas we build for.
Each row is a research-use or business-analysis workflow and the analysis infrastructure under it. GenomeRx does not provide clinical, diagnostic or medical services, and it does not promise research outcomes.
- Bioinformatics and data analysis
- Pipelines for sequencing and other research data, with quality control, statistics and reporting as maintained code.
- Drug-discovery research
- Data preparation, model training and inference, and analysis of screening data, on compute sized for those runs.
- Biomarker research
- Pipelines that carry cohort and omics data through to exploratory analysis, built so every result can be rerun.
- Pharmacogenomics research
- Research-use genomic workflows, such as variant processing and annotation, with documented inputs and outputs.
- Analytics, AI and data automation
- Reproducible data analysis and reporting for research and business questions, and extraction, classification, summarization and search over internal documents and data, with review points for people. Each run is measured for runtime and resource use, outputs carry quality checks, and the operational metrics to watch are agreed in the scope.
Tested on the built system, then handed over.
Pre-deployment testing is acceptance testing: the workflow and the hardware are checked against criteria the client agreed in the scope. It is not a certification or a regulatory approval.
- Acceptance testing Delivery stage 04
- The workflow runs end to end on the built hardware, in the environment where it will live. Outputs, runtime and resource use are compared with the agreed criteria, and a rerun has to reproduce the result. A failed test goes back to the workflow or the build.
- Handover and support Delivery stage 05
- The code repository, the environment and build specification, runbooks and the test record are handed over with the system, with a walkthrough for the people who will run it. Support continues under a monthly retainer if wanted.
Tell us what needs to run.
Send a short note on the question or process, the data and its rough size, where the system has to run, and your timeline. We reply by email to arrange a scoping call.
Engagements
GenomeRx sells services to organizations: architecture, workflow code, hardware and compute integration, acceptance testing, documentation and support. Hardware is quoted as part of the scoped system, not sold separately.
- Evaluation sprint
- We evaluate a candidate workflow or model on your task and measure the compute it needs. Fixed fee.
- Workflow or system build
- Architecture, code, hardware integration and acceptance testing. Milestones under a statement of work.
- Integrated delivery
- The build and the workstation, appliance or server it runs on, as one scoped system. Services plus quoted hardware.
- Support
- Help after handover for the people who run the system. Monthly retainer.
B2B only, scoped in a written agreement. Client data, code and project details stay confidential and are not published here.