Talaria Scientific

Optimal design, in silico first

For teams whose designs are built and tested: a material, a component, a process, a recipe.

Hosted TauOpt: closed beta todayTalaria hosted: alpha December 5, invite onlyEnterprise: coming soon

An optimal design problem is a geometry, a material, a process setting, or a recipe that must meet objectives within constraints, when every evaluation by simulation or physical test is costly. Talaria Scientific makes the design loop, from candidate to test and back, fast, trustworthy, and cheap, so you can try orders of magnitude more candidates in silico before you build one.

Talaria builds surrogate models from your own ideas, data, and simulations: validated stand-ins for a simulation or a physical test that answer fast enough to search over, and that are yours. TauOpt searches over them, and over any other evaluator you have, and returns the short list to test. The results flow back, the models update, and the number of physical tests to reach a target falls.

Today you can start with hosted TauOpt over your own evaluators, by request. The Talaria hosted alpha on December 5 adds the models. The enterprise deployment brings the whole loop into your cloud.

Contact us →Request TauOpt access
Accelerated modeling, optimization, and design: your ideas or data become validated digital twins in the Talaria Harness, the TauOpt Optimizer searches over the twins, your other evaluators, and the results that feed back, and the best result improves run by run, all of it kept in the record ACCELERATED MODELING, OPTIMIZATION, AND DESIGN THE RECORD every candidate, run, result, and decision in here is kept with its provenance 1 Your ideas or data a hypothesis, a model to try, simulations, process data 2 Talaria Harness builds validated digital twins from your ideas and data 3 Digital twins surrogates, reduced-order models, tuned solvers 6 Results output metrics of process to optimize 5 Candidates the next configurations to evaluate 4 TauOpt Optimizer picks the next candidates to test Your other evaluators simulations, backtests, physics foundation models 7 Best so far objective iterations ideas, data builds and validates predictions queries proposes to test results, fed back into the search queries results the objective, run by run create or update models = a product (what we build) = an artifact of the loop = yours: ideas, data, evaluators, results, the twins = the record = the output 1 = a step in the loop = a return path: create or update models
Figure 1.The same loop, for a design problem: your ideas and process data go in; the Talaria Harness turns them into digital twins; the TauOpt Optimizer searches over the twins, your other evaluators, and the results that feed back; the best result improves run by run; the candidates, runs, and results land in the record.

Two stages and a return path

The loop has an in-silico stage, a physical stage, and a return path that feeds each cycle's results into the models.

  1. In silico. TauOpt explores the design space over the surrogate models and returns a short list of candidates with the highest expected value.
  2. Physical. You build or test the candidates.
  3. Back. The results return: the surrogates update with them, and TauOpt proposes the next batch.

How to get it

Read:The scientific flywheel that can lead to abundance (where the search is going)Surfing the Frontier Wave that rests between The Bitter Lesson and No Free Lunch

Get access, or get in touch

Talaria is in private beta with a small group of researchers, and TauOpt is in closed beta. For product access, use the request forms. For enterprise, partnerships, or press, use the contact page.

Contact us →Request Talaria accessRequest TauOpt access

Talaria: the architectureTalaria: the blogTauOpt: documentation

100%Download SVG