Tesryn Lab
What we've learned to build.
Engineering automation first: cells, CAD, parts and the figures behind them, checked against the makers’ data, most of it now Aximyr. Consulting compounds: what we build for you is yours, and the know-how is what we carry to the next client. If none of it is your problem, yours is next.
- 14things we can build
- 3on the bench
- 9kinds of work
01 — The catalogue
What kind of work is yours?
Pick one and see only that. Nothing here is software to buy — it's what we've already learned to build, and yours would be your own.
Tesryn Signal
Scan a code, describe a problem, and watch the system that solves it take shape.
First built for Tesryn, at events and in meetingsPodcast Maker
A private podcast that teaches one listener what they don't know yet, from sources they trust.
First built for a single listener who learns by earMagnetic Digital Twin
Design a magnetic instrument on screen and know, with evidence, which version is better.
First built for a laboratory instrument research programCAD Workbench
Put real parts together on screen and know, before anything is ordered, whether they fit.
First built for engineers who have to know a design fits before it is madeEngineering Library
Ask the papers and datasheets you already own, and get the passage that answers, with the page it came from.
First built for engineers whose answers are buried in papers they already haveReconstruction Research Studio
A research workbench for breast reconstruction after mastectomy: comparable cases, plain explanations, and room for clinicians to steer it.
First built for a clinical research collaborationPart Finder
An inside sales rep types what a customer asked for and gets the part, the equivalents, the nearest stock and a quote.
First built for an industrial distributor's inside sales teamApplication Engineering
From what a customer needs a machine to do, to a first design package: parts, checks, a bill of materials and a model.
First built for an industrial prospectReferral Engine
Match each enquiry to the right practitioner, with the reasons written down.
First built for a network of independent practitionersCommand Center
A founder describes a change to the company, and sees it drafted, reviewed and published.
First built for a founder running a growing companyResearch Benchmark
Know whether a change to an AI research assistant actually made it better.
First built for teams improving AI research and search toolsCapability Distillation
When an AI figures something out, decide where that know-how should live for good, and prove the change helped.
First built for teams paying for the same AI work twiceSentinel
Watches a machine that runs AI, explains what changed, and never touches a thing.
First built for workstations and servers that run AI around the clockTesryn Launch
The scope a customer agreed to, tracked to delivered and accepted, with the evidence attached.
First built for anyone delivering agreed work to a customerReferral Platform
Offer a referral to the best-matched practitioner, and cascade it if they can't take it.
For a network of independent practitionersInsurance-aware Intake
Ask about coverage at the start, so nobody is matched to someone they can't afford.
For networks whose clients claim on insuranceResearch Runtime
An AI researcher that has to prove it's better than the last one before it ships.
For teams that need answers with sources behind themNothing built in that kind of work yet — which is usually where a project starts. Tell us about yours.
02 — On the bench
3 more, not finished, and marked as such.
Being built now, or designed and waiting. If one is what you need, say so — it moves up the list and you help shape it.
- Launching soon Referral Platform Offer a referral to the best-matched practitioner, and cascade it if they can't take it. For a network of independent practitioners
- In the workshop Insurance-aware Intake Ask about coverage at the start, so nobody is matched to someone they can't afford. For networks whose clients claim on insurance
- In the workshop Research Runtime An AI researcher that has to prove it's better than the last one before it ships. For teams that need answers with sources behind them
03 — The map
What we know, and what proves it.
Each entry is the visible part. Underneath, the same handful of things we have learned to do well keep coming back in different industries, and every line here means one of them was proved by real work. It is also why an unfamiliar problem is not a blank page for us.
Evidence & trust
Making a claim checkable, and keeping the record honest.
- Sealed, verifiable records5
Work that writes down what happened in a way nobody can quietly edit later.
Proved by Podcast Maker, Magnetic Digital Twin, CAD Workbench, Sentinel, Tesryn Launch.
- Every claim traced to its source6
Nothing is asserted without the passage, part number or measurement behind it.
Proved by Podcast Maker, Engineering Library, Reconstruction Research Studio, Part Finder, Application Engineering, Research Benchmark. Being built: Research Runtime.
- A person approves before it's real7
Automation can prepare anything and finish nothing. The last step is always a person.
Proved by Tesryn Signal, Podcast Maker, CAD Workbench, Reconstruction Research Studio, Application Engineering, Command Center, Tesryn Launch.
Judgement under uncertainty
Knowing what the numbers can and can't tell you.
- Comparing only where evidence can tell things apart3
Differences smaller than what the method can measure are reported as ties, not wins.
Proved by Magnetic Digital Twin, Research Benchmark, Capability Distillation.
- Knowing whether a change actually helped2
Conditions written before the run, a sealed test, and a verdict that stands either way.
Proved by Research Benchmark, Capability Distillation. Being built: Research Runtime.
- Saying how sure it is, in plain words2
Confidence expressed so a non-specialist can weigh it, not hidden in a number.
Proved by Magnetic Digital Twin, Reconstruction Research Studio.
Finding and matching
Getting the right thing, or the right person, in front of someone.
- Matching a person can check2
Weights, hard requirements and capacity — arithmetic with its reasons attached, never a model's hunch.
Proved by Part Finder, Referral Engine. Being built: Referral Platform, Insurance-aware Intake.
- Search that explains itself3
Loose human wording in, ranked results out, each saying what matched and what differs.
Proved by Engineering Library, Part Finder, Application Engineering.
Physical systems
Simulating the world well enough to decide with.
- Simulation you can decide with2
Fields, forces, heat and soft tissue, solved fast enough to explore and checked against known answers.
Proved by Magnetic Digital Twin, Reconstruction Research Studio.
- A description becomes real parts3
Numbers in, manufacturable geometry and engineering limits out, the same way every time.
Proved by Magnetic Digital Twin, CAD Workbench, Application Engineering.
Systems people can drive
Interfaces someone non-technical runs without training.
- Plain language, no training day6
One action per screen, words the user already uses, and errors that say what to do next.
Proved by Tesryn Signal, Reconstruction Research Studio, Part Finder, Referral Engine, Command Center, Tesryn Launch. Being built: Referral Platform, Insurance-aware Intake.
- Say it, see it, approve it, publish it3
Changes described in words become structured edits with an exact preview and an exact undo.
Proved by Tesryn Signal, Referral Engine, Command Center.
- One conversation, many screens1
A phone, a big screen and a tablet showing the same moment, live, without anyone operating them.
Proved by Tesryn Signal.
Running it safely
Where it runs, what it keeps, what it may touch.
- Runs on hardware you control4
Models and data that never have to leave the building when the material is sensitive.
Proved by Podcast Maker, Engineering Library, Command Center, Sentinel. Being built: Research Runtime.
- Watching without touching2
Systems that can detect and explain but hold no power to change what they watch.
Proved by Capability Distillation, Sentinel.
- Only what's needed, deleted on request2
The least data that makes the thing work, with forgetting built in rather than promised.
Proved by Tesryn Signal, Referral Engine. Being built: Referral Platform, Insurance-aware Intake.
If none of these is your problem
Then yours becomes the next thing we learn.
Every skill here started as somebody's ordinary problem, and none of them existed before they asked. This isn't a shelf of software to buy: you describe what's slow, wrong or impossible today, we build it for you, and you own it outright. We come out of it able to do it again — which is exactly what the client before you paid for, and why yours goes faster.
Not sure it's possible? That's a good first question and it's free to ask. We'll say plainly what we'd build, what we wouldn't, and what it would cost.