Who this is for
This is for small teams considering local predictions, semantic search, retrieval-augmented generation, or a more controlled AI-assisted development workflow. We first check whether simpler software would solve the problem.
Costs and ownership
Our research repositories are publicly available under their respective licences. Advice, integration, and support can be paid engagements. We agree scope and fees before starting; third-party model licences and infrastructure costs are assessed separately.
How we work together
- Define the decision to improve, available data, privacy constraints, and a baseline to compare against.
- Build a limited prototype and evaluate it on representative examples, including failures.
- Document results, limitations, handover, and whether further integration is justified.
See the work
MokuML explores local predictions from Postgres data. BlackBoxApproach explores human decisions and module boundaries in AI-assisted development. Both are research proofs of concept, not evidence of production outcomes or guaranteed savings.
Support and boundaries
Feasibility depends on data quality, hardware, model limits, and licences. We explain what stays local and what would be sent to an external provider before choosing an architecture. A prototype is evaluated separately from a production deployment.
Discuss your project
Describe your workflow, the problem you want to solve, and the devices or data involved. Please leave passwords and private customer data out of your message.