Services
The kit on GitHub is the architecture and a 944K research-demo model. Putting either of those into a shipping product takes silicon bring-up, signed boot, domain fine-tuning, and adversarial hardening. We do those — fixed scope, fixed price, source artifacts delivered to your control.
Prices below are starting points for the scope described. To start: email hello@atomelm.com with a one-paragraph brief — your hardware target, your domain, what you want the model to do. No phone, no calendar links, no calls.
.atome blobs bound to device UID, dev/prod flag separation, per-platform secure-boot integration (STM32 Trust Zone, RP2350 secure boot, ESP32-S3 flash encryption), remote attestation, tamper-evident audit log of every inference. Includes threat-model writeup + a fuzz harness on your input grammar.
ATOME01, and deliver the C-engine-ready blob plus the PyTorch checkpoint plus training logs. If the corpus is too small or too noisy we tell you before we start.
ATOME01 blob, and an annual refresh.
How we engage
Email hello@atomelm.com with your hardware target, your domain, and what you want the model to do. That's it — no contact form, no calendar link, no call.
Within 48 hours you get a one-page scope document and a fixed price, by email. Reply with sign-off — or not. No follow-ups.
We sign your NDA before any data lands on our hardware. Source artifacts delivered to your control — no DRM, no per-device fees on the open-source surface.
Fixed-scope, fixed-price unless explicitly retainer-based. Source artifacts delivered to your control — no DRM, no per-device fees on the open-source surface. NDA before data. Email-only. No phone, no calendar invites, no calls — every inquiry handled in writing.