R-01
Retrieval that holds at corpus scale
Most quality loss happens at ingestion, not inference. Chunking, hybrid scoring and freshness decay.
Active
Research lab
A lab with a delivery obligation.
Every thread here opened because a client system was not good enough. Findings land in a product within a quarter or two.
How the lab works
Three rules that keep research honest.
Problems come from production
A thread opens when a real deployment hits a ceiling: quality, latency, thermal envelope, reliability.
Findings must be reproducible
If a delivery engineer cannot apply it within a sprint, it is not finished research.
Negative results are kept
The approaches that did not work are a small lab's most valuable asset, and we tell clients about them.
Active threads
What we are working on right now.
R-02
Distillation for constrained hardware
Task-specific capability compressed into the power and thermal envelope of a board we can build.
Active
R-03
Document structure understanding
Reading order, nested tables, stamps and handwriting — a page as spatial structure, not a token stream.
Active
R-04
Agent reliability under real tools
How multi-step agents fail when tools are slow or wrong, and which supervision patterns recover.
Active
R-05
Evaluation as a first-class artifact
The eval set as the deliverable clients keep: cheap, honest, adversarial suites that catch regressions.
Ongoing
Research summaries are illustrative descriptions of internal work; experiment notes are shared under NDA.
Hardware co-design
The lab has a machine shop behind it.
Distillation targets are chosen against a board Xelec can manufacture, and validated on the bench.
5
Active research threads
< 1 qtr
Typical finding to product
60k
Frames in a distillation set
2
Reference boards in validation
Next step
Have a problem that needs research, not integration?
If your use case sits beyond what off-the-shelf models do, that is the work the lab exists for.