← Semiconductors board · written 2026-10-18 by claude-opus-5-5 from the pipeline's evidence only

Tai An (Anai-Guo): semiconductors, Skoor 35 (n=3, confidence 0.7)

What they built

No owned project appears in the evidence. Tai An is listed as a committer on three repositories owned by others [source].

k2-fsa/sherpa-onnx: offline speech recognition and synthesis on onnxruntime, with support for embedded systems, RISC-V, and RK, Axera and Ascend NPUs. It has 15,339 stars and its primary language is C++ [source].
fastmachinelearning/hls4ml: machine learning on FPGAs using HLS. It has 2,189 stars [source].
RLinf/RLinf: reinforcement learning infrastructure for embodied and agentic AI. It has 5,430 stars and is tagged robotics [source].

What their commits change in these repositories is not in the evidence.

Record

These are the measured signals.

The profile shows 549 commit contributions in the last year [source].
The account was created 2026-04-12 [source].
They committed to 2 semiconductors-tagged repositories in the last 90 days: sherpa-onnx and hls4ml [source] [source].
The profile has 20 followers [source].
Company, location, bio and links are blank [source].
Papers are not in the evidence.

The per-repository 90-day commit counts (135, 55 and 154) are not attributed to this person in the evidence [source].

Inferences:

Inference: 549 contributions in about six months is high volume, but contribution counts do not measure size or difficulty.
Inference: sherpa-onnx is speech software that runs on NPUs. The "semiconductors" tag rests on hardware targets, not chip design.
Inference: hls4ml is the closest fit to the semiconductors thesis.
Inference: a new account with high output may belong to someone who previously worked under another account or inside a company.

Why now

The only timing evidence is a new account that was active across all three repositories in the last 90 days [source]. Market or company timing is not in the evidence.

Unknowns

Identity beyond the name "Tai An" is not in the evidence [source].

The scope of their commits and any maintainer status are not in the evidence.

Employer and any affiliation with k2-fsa, RLinf or hls4ml are not in the evidence.

Any company or plan to found one is not in the evidence.

Prior GitHub history before 2026-04 is not in the evidence [source].

Questions for the room

What did your hls4ml commits change? Did any touch FPGA synthesis or hardware backends, or were they tooling and docs? This tests whether the semiconductors tag reflects hardware work.

Your account is from April 2026, but you have 549 contributions since. Where did you write code before, and under what name or employer? This tests the new-account signal.

Are you paid to work on sherpa-onnx or RLinf, and by whom? This tests whether the work is independent or employer-directed.

What would move the Skoor

The Skoor would rise with merged hls4ml commits to hardware backends, maintainer status, or a founder-owned repository in semiconductors. It would fall if the commits are mostly small fixes or docs, or if the work is employer-assigned.