What we doWorkPeopleConsultation
We analyze your business, recommend what fits it, and implement the choice end to end.
Consultation01 · ADVISORY
Business needs, budget, constraints, the nature of the business. Then a concrete recommendation: which infrastructure, which tooling, which models.
02 · IMPLEMENTATION
We build and deploy what we recommend. Hardware, software, complete working systems that keep working after we leave the room.
03 · COMPUTE
Run and train your models on our machines. Capacity sized to the workload, direct access to the engineers who run them, terms agreed one to one.
Three recent engagements. Client names stay private. The work does not.
The client needed AI capability without sending data outside the building. We analyzed their workloads and designed the full local setup: servers sized to the load, worker fleet, serving stack, and the operating procedures to run it.
We set up Hermes agents for a production environment: tools and permissions scoped per agent, monitoring from day one, and human approval gates wherever a decision carries risk.
We implemented OpenClaw running inference locally through custom-trained models served on vLLM. Every token generated on the client's own hardware.
Every engagement starts the same way: a private consultation, free, with the person who does the work.
Three recent engagements. Client names stay private. The work does not.
The client needed AI capability without sending data outside the building. We analyzed their workloads and designed the full local setup: servers sized to the load, worker fleet, serving stack, and the operating procedures to run it.
We set up Hermes agents for a production environment: tools and permissions scoped per agent, monitoring from day one, and human approval gates wherever a decision carries risk.
We implemented OpenClaw running inference locally through custom-trained models served on vLLM. Every token generated on the client's own hardware.
Every engagement starts the same way: a private consultation, free, with the person who does the work.
Needs, limitations, vision. We ask until we understand the company, not just its systems.
Workloads, data flows, budget, constraints. What actually runs, and what it costs to run it.
Hardware, tooling, and models, with reasons you can check and a price fixed before work begins.
Machines racked, software configured, systems deployed to production. End to end, by the same people.
We stay reachable, and we say so in writing.
Needs, limitations, vision. We ask until we understand the company, not just its systems.
Workloads, data flows, budget, constraints. What actually runs, and what it costs to run it.
Hardware, tooling, and models, with reasons you can check and a price fixed before work begins.
Machines racked, software configured, systems deployed to production. End to end, by the same people.
We stay reachable, and we say so in writing.
Tools and platforms we set up and operate as a matter of daily work. No certification, partnership, or endorsement implied.
Michal Sawosz designs and runs AI systems end to end: hardware on the floor, models in production, agents that hold up under real work. Your consultation is with him. So is the build.
MICHAL SAWOSZ · FOUNDER, ASTRA TECHNOLOGIES
Michal Sawosz designs and runs AI systems end to end: hardware on the floor, models in production, agents that hold up under real work. Your consultation is with him. So is the build.
MICHAL SAWOSZ · FOUNDER, ASTRA TECHNOLOGIES
Every project starts with a free, private consultation. Tell us a little about your company below; it lets us prepare, and it makes the first hour count.
No pricing pages, no calculators. Scope decides cost, and scope is a conversation.
Prefer to talk? +48 500 100 120
