Services
Senior consultants on your hardest problems, by the hour or by the statement of work.
GPU infrastructure, healthcare data, and research are a sample of where our consultants have shipped production work. Between them, the twelve cover ten sectors. Every engagement starts with thirty minutes with Shaun on one problem you name.
- GPU infrastructureObservability for GPU infrastructure, built end to end
- Healthcare dataInside your governed environment, by the hour or by SOW
- Research and deep techQuantum, HPC, and life-science researchers who also write the code
- EvaluationsEvaluation suites built for your workflows, scored over the whole agent
We hold no HIPAA attestation of our own — we work inside your governed environment, under your BAA.
Highlights
A selection of the work
A sample, not the limit. Each of these has its own page; the sectors below and the team profiles give the full picture.
GPU infrastructure
InfrastructureObservability for GPU infrastructure, built end to end, with scheduler, GPU, and node telemetry, fabric, storage, tracing, and logs in one multi-tenant system. Finance-grade systems with security built into the architecture.
Healthcare data
Healthcare dataAnalytics teams stood up inside large regulated environments, and evaluation frameworks built at platform scale. Usually a short scoped SOW on your own data, in your environment.
Research and deep tech
ResearchQuantum computing, high-performance computing, and life-science research, from consultants who also write the code. Usually a part-time principal, by the hour.
Employers named on this site are where our consultants worked, not Transmute Data clients.
More sectors
Where else our consultants have shipped
Name the problem on the first call and we'll say who has worked it before.
Financial services
Quantitative research and risk at a global macro fund, production ML at a commercial lending platform, money-laundering detection, and an LLM multi-agent anti-money-laundering prototype.
Platform integrity
Seven-plus years at Twitter and X designing graph-based coordinated-behavior detection, then quality measurement infrastructure at Meta (C-07).
Agentic systems
Multi-agent workflows, retrieval pipelines, and a production conversational agent, plus moving an engineering org to shipping with agents daily.
The electric grid
Demand and generation forecasting, AMI and SCADA, DER adoption modeling, and lead development on a state electrification impact study.
Defense and autonomy
Predictive-maintenance models for US Air Force E-3 Sentry aircraft, and formally verified control for robots and autonomous vehicles.
Security and consumer products
Patents in least-privilege security (the individuals'), security built into finance-grade systems, and staff-level consumer mobile engineering.
Each entry is one of our consultants' own work, in the roles and at the employers where they did it. Prototypes and trials are labeled as such. Meet the consultants
Engagement options
Two ways to work with us, one way to start
Every engagement starts with the same conversation, then takes one of two shapes. Handover is in scope either way.
Start · First conversation
A 30-minute call
With Shaun, on one problem you name. We come back with the consultants who have already solved that shape of problem before, with their availability to meet.
- Shaun takes the first call
- Listed consultants who have solved it before
- Next steps, meeting with the named consultants
Shape 1 · Forward-deployed
Senior consultants, by the hour
Named senior consultants embedded in your team, by the week or month, full- or part-time, extended in place.
- Suits teams that know the problem and need the person
- Your code review, your CI, your definition of done
- Start part-time and extend as the work grows
Shape 2 · Statement of work
A scoped, working outcome
A working outcome you can check, within a defined window, scoped around the part your team can't staff today. It often starts small, on your own data, in your environment.
- Suits a team that needs a fixed scope and a defined window
- If the data will not carry it, we say so early and in writing
- Handover is in scope, not a follow-on quote
We don't publish rates. We quote each engagement after the first call, against the scope you name.
Who shows up
A principal leads the work and stays on it
Every engagement is led by a principal-level data scientist who stays through delivery. An AI engineer owns architecture, build, and production, and a delivery lead handles coordination and stakeholder alignment. Other specialists from the twelve join as the scope needs them.
On a small or research-shaped problem, the principal may be the whole team, part-time. You meet the person accountable for delivery before the work starts.
How an engagement runs
Name the problem. Meet the person. Build in your stack. Hand it over.
Name the problem
A 30-minute call with Shaun on one problem you name.
Meet the person
Shaun names who has already solved that shape of problem before. You meet them before the work starts.
Build in your stack
In your repositories and tooling, through your code review and CI, carried to production. We work to provide value as quickly as possible.
Hand it over
Documentation, runbooks, monitoring thresholds, and the evaluation set. You own the deliverables once they're paid for. We keep the tools and frameworks we brought with us.
Security and data handling
In your environment, under your access
Our consultants work in your repositories and environments, under the access your security team grants. Before anyone touches your data, we sign a confidentiality agreement, and access is scoped to the people staffed and ends when they roll off.
We build our own designs, agents, harnesses, skills, and processes, and improve them using our own internal development work. Never customer data, never PHI. We hold no HIPAA attestation of our own — we work inside your governed environment, under your BAA.
You own the deliverables once they're paid for. We keep the tools and frameworks we brought with us. Work delivered through a partner follows that partner's agreement.
A 30-minute call, on one problem you name.
Shaun takes the call. You leave knowing which shape fits, hourly or SOW, and who in the practice would lead the work.