KhazLabz

🔩

Firmware & Hardware

Embedded systems taken from an idea on a napkin to a device that survives in the field, with the firmware written to be maintained rather than rewritten.

Schematic review, board bring-up and hardware debugging
Firmware in C/C++, Rust and MicroPython for ARM Cortex-M, ESP32 and RISC-V
Drivers, RTOS integration and low-power design
Sensor integration, fieldbus and wireless (I2C/SPI/UART, Modbus, BLE, Wi-Fi, LoRa)
Secure OTA updates and device provisioning

Let's talk

🧪

Test engineering

Six years of my day job, in automotive. Anything from a single board to a whole system, tested so that regressions show up on the bench rather than in the field.

Test strategy and requirements-based test design (ISTQB certified)
Hardware-in-the-loop rigs and automated bench setups
map[Automotive tooling:Vector CANoe, Lauterbach TRACE32, CAN/LIN/Ethernet buses]
Test automation in Python and CI integration for firmware and backend alike
Evaluation harnesses for LLM and RAG pipelines, so "it works" is a number, not a feeling

Let's talk

⚙️

Backend & Infrastructure

Services that talk to those devices and to your users, plus the infrastructure they run on. I prefer boring, self-hostable technology and I write the runbook.

Backend services in Go and Python (FastAPI), APIs and integrations
Server-rendered web apps with HTMX, without a frontend framework tax
PostgreSQL and SQLite design, migrations and performance tuning
Containers with Podman/Docker, Kubernetes where it actually pays off
Infrastructure as Code with OpenTofu and Ansible, on-prem or EU cloud (Scaleway, Hetzner)
CI/CD, monitoring and observability (Prometheus, Grafana, OpenTelemetry)

Let's talk

🧠

AI infrastructure & pipelines

The sprinkle on top. Not chatbots for the sake of chatbots, but the plumbing that makes a model useful and keeps your data where it belongs.

Self-hosted LLM inference (LocalAI, vLLM) behind a LiteLLM gateway, on your own or EU-based GPUs
RAG pipelines with open models such as Bielik, including evaluation
Data pipelines that feed models and dashboards, from device telemetry to warehouse
Agent and MCP integrations wired into existing systems
Honest assessment of whether you need a model at all

Let's talk

How we can work together

Hourly / daily

Debugging sessions, reviews, audits, and anything under two weeks.

Fixed scope

A defined deliverable with milestones, ideal for a board bring-up or a service rewrite.

Retainer

A fixed number of days per month for ongoing maintenance and on-call for firmware and infrastructure.

Why KhazLabz?

You talk to the engineer

There is no account manager. The person who scopes the work is the person who writes the code and answers the phone when it breaks.

Whole-stack view

Most bugs live at boundaries. Owning the whole stack, from the hardware up to the web frontend, means I can find them there instead of passing tickets around.

Open source by default

My own work ships under GPL/AGPLv3 for software and CERN-OHL-S for hardware (with a commercial licence when you need one), or Apache 2.0 for smaller projects. I use open tooling and document what I build so you can take it elsewhere.

European and self-hosted

Data stays in the EU, preferably on infrastructure you control. I will not lock you into a vendor I would not use myself.

Have a device, a service or a pipeline that needs an owner?

Send me a short description of the problem. I reply within two working days with questions, not a quote template.

Get in touch