I'm Bayar — I build AI infrastructure in public

I'm a software developer focused on AI infrastructure — the plumbing that sits between a product and a model. I like the unglamorous parts: streaming transport, validation, rate limiting, failover, and the guardrails that decide what reaches the model and what comes back.

I build in public because it keeps me honest and because the work is more useful when someone can read exactly how it fits together. The live terminal on the home page is a real piece of that work, not a mockup — it runs on an endpoint I wrote, documented in full on the docs page.

const bayar = {
  building: "bayar.dev — AI infrastructure, in public",
  stage:    "early, bootstrapped, shipping",
  focus:    ["streaming AI", "edge routing", "rate limiting",
             "semantic caching", "PII redaction", "private VPC / BYOK"],
  daily:    "CachyOS + Niri, kitty + fish, too many terminals",
  based:    "Walnut Creek, CA",
};

Three layers of the stack hold my attention. The first is live today; the others are where the work is heading.

The interactive terminal on the home page connects to a real streaming endpoint with request validation, per-identity rate limiting, and server-side error isolation. Its reference-architecture sections describe where I'm taking it — edge caching and private-cloud deployment — and are labelled as design, not as a deployed system. The only performance figures on this site are the ones on /benchmarks, measured every 15 minutes with the raw rows published. The request and response contracts are written up in the technical documentation.

• GitHubWhere I build in the open@namchinbayar ↗
• LinkedInBackground and work historyin/bayardev ↗
• XShort updates from the build@bayardotdev ↗
• YouTubeOccasional video@bayardotdev ↗

Email hi@bayar.dev. I read everything and reply to what I can.