optim · engineering

Contract engineering,
in named areas.

Six areas we are held to, each with the standards we work against and a link to the work that evidences it. Most engagements start as one area and grow into the ones next to it, because the team writing the firmware is the team operating the control plane your fleet reports to.

──── [01] service areas ──── ────
01 // wireless-firmware

Wireless firmware & bring-up

Access points and gateways from board bring-up to the OTA path that keeps them maintainable, across MediaTek Filogic, Qualcomm IPQ and Broadcom BCM.

  • Board bring-up, BSP and driver work on vendor SDKs
  • 802.11ax / 802.11be radio bring-up, tuning and AFC
  • EasyMesh controller and agent, roaming and band steering
  • Signed images, dm-verity root and the update path
OpenWrt · OpenWiFi · prplOS · RDK-B
Read the carrier firmware bring-up →
02 // device-management

Device management & ACS

TR-069 and TR-369/USP control planes that survive a real fleet: multi-tenant by MSO, health-gated campaigns, and an edition you can host yourself.

  • TR-069 / TR-369 (USP 1.3) agents, controllers and data models
  • Multi-tenant fleet control with per-MSO isolation
  • Health-gated firmware campaigns with automatic rollback
  • Single-binary self-hosted edition — Postgres and nothing else
TR-069 · TR-369 / USP · TR-181 · EST
Read the global ACS case study →
03 // platform-integration

Platform integration & migration

Moving an estate without a flag day: bringing an existing fleet or workload onto infrastructure you control, with the identity and network plumbing that has to work first.

  • ACS migration with the fleet online
  • Workload moves off hyperscalers into a jurisdiction you name
  • Private PKI and OIDC — step-ca, Authentik
  • BGP, anycast and multi-tenant network design on our own ASN
AS207819 · Talos · Cilium · step-ca
See the infrastructure domains →
04 // embedded-linux

Embedded Linux & BSP

Kernel, drivers and build systems on ARM and MIPS targets — the layer underneath everything else, kept reproducible so the next person can rebuild it.

  • Kernel and device-driver development, upstream and backport
  • Yocto and Buildroot BSP layers locked by revision
  • Bootloader, secure boot and root filesystem integrity
  • Per-SKU SBOM and a CVE response path that holds up in audit
Yocto · Buildroot · dm-verity · CycloneDX
Check the published components →
05 // camera-and-sensing

Camera firmware & sensing

IP cameras, ANPR and purpose-built optics: ISP tuning that holds in evening light, multi-stream encoders that do not drop frames under WAN jitter.

  • ISP 3A tuning, WDR / HLC curves per SKU
  • H.264 / H.265 / AV1 hardware encode, multi-stream
  • ONVIF Profile S / T / M conformance, RTSP over TCP + UDP
  • On-device models — detection, ANPR, face blur for GDPR
ONVIF · RTSP · SRT · WebRTC
See the camera firmware section →
06 // sustaining-engineering

Validation & sustaining engineering

The years after launch. Long-lived device software needs someone who still understands it in year five, and a validation record that says what was measured and what was withdrawn.

  • Conformance and interoperability testing on the actual SKU
  • Regression suites wired into CI, hardware-in-the-loop
  • CVE response, kernel and toolchain uplift, silicon EOL moves
  • Claim auditing: copy linted against a ledger of what we measured
TR-069 / USP · EasyMesh · ONVIF · claim-lint
Read the validation record →
──── [02] engagement models ──── ────

Fixed-price project

Defined scope, deliverables and a date. We scope it on a call, write the statement of work with named engineers, and the price does not move unless the scope does.

  • ▸ Written scope and acceptance criteria
  • ▸ Fixed deliverable and date
  • ▸ Change requests priced separately

Time & materials

For work whose scope cannot honestly be fixed up front — bring-up on new silicon, a migration with unknown fleet state, or investigating something that is not reproducible yet.

  • ▸ Monthly billing against timesheets
  • ▸ Weekly demos and a visible burn
  • ▸ Stop or redirect at any month boundary

Dedicated team

A team that works as part of your engineering organisation, on your tools and your standups. This is how most of our long-running firmware programmes run.

  • ▸ Named engineers, not a shared pool
  • ▸ Your process, your ticket system
  • ▸ Shared channel with the engineers doing the work