Consulting & advisory

An outside read on the
hardest part of your program.

Most autonomy efforts don't fail on the algorithm. They fail on architecture decisions made eighteen months earlier, on a link budget nobody owned, or on a subsystem that was supposed to be a two-week integration. I've been on the wrong side of all three — and shipped anyway.

Start an inquiry What I take on

Engagements

Three ways I'm useful

Scoped tightly, delivered with something written down at the end. No open-ended retainers that quietly become headcount.

01

Architectural reviews

A structured, independent review of an autonomy or robotics system architecture — what will hold at scale, what is quietly load-bearing, and what will cost you a rewrite in a year. Useful before a platform commitment, ahead of a funding or program milestone, or when a system that worked in the demo has stopped scaling.

  • Compute, sensing, comms and software topology assessed end to end
  • Failure-mode walkthrough: degraded links, denied navigation, partial sensor loss
  • Build-vs-buy and standardization calls, with the cost of each written down
  • Deliverable: a findings document with prioritized, sequenced recommendations
02

Autonomy systems development guidance

Hands-on technical direction for teams building autonomous vehicles — ground, air or subsea. This is the work of turning a promising capability into a program with a credible path to the field: sequencing, technical strategy, and the judgment calls that are expensive to get wrong.

  • Roadmap and milestone definition tied to actual field readiness, not demo readiness
  • Autonomy stack design: state estimation, localization in denied environments, decision pipelines
  • Vehicle health, readiness automation and the test infrastructure that keeps a fleet flyable
  • Team-facing: design reviews, mentoring senior engineers, unblocking specific decisions
03

Specialized subsystems & integrations

Delivery work on the pieces that are genuinely hard and rarely worth hiring for permanently. I build the subsystem, integrate it against your stack, and hand it over with the documentation and tests to own it.

  • Low-latency, high-bandwidth video and control links; telemetry that degrades gracefully
  • Edge compute bring-up, accelerated inference, and embedded platform standardization
  • Sensor suite integration and calibration — camera, LiDAR, acoustic, inertial
  • Firmware across distributed microcontrollers; legacy avionics and controls modernization
  • Turning raw sensor output into a customer-grade deliverable

How it works

Short path from question to answer

  1. Intro call — 30 minutes, no charge You describe the problem. I tell you honestly whether I'm the right person, and if I'm not, what kind of person is.
  2. Scope & proposal A written scope: the question being answered, the deliverable, the timeline and the fee. Fixed-scope wherever it can be.
  3. The work Reviews typically run one to three weeks. Guidance engagements run on a recurring cadence. Subsystem builds are scoped per project.
  4. Handover Findings written down, decisions justified, code documented and tested. Your team should be able to carry it without me.

Remote by default, on-site where the problem is physical — and in this line of work it usually is. Available for both commercial and defense-adjacent programs.

Good fit

  • Robotics or autonomy programs heading toward real deployment
  • Heavy industry, energy, maritime, subsea and defense
  • Teams with hardware in the loop and a date they have to hit
  • Organizations standardizing edge compute or embedded platforms
  • Founders who need an experienced second opinion before committing

Probably not a fit

  • Pure web or app development with no physical system involved
  • Staff augmentation billed by the hour with no defined outcome
  • Work that needs a full-time engineer rather than a specialist

Inquiries

Tell me what you're building

The more specific the problem, the more useful the first call is. A few sentences on the system and where it's stuck is plenty.

Send an inquiry

One email is enough to start. The more specific the problem, the more useful the first call is — a few sentences on the system and where it’s stuck is plenty.

  • What you’re building, and the domain it operates in
  • Where it’s stuck — the decision or blocker in front of you
  • Any timeline you’re working against
  • Which of the three engagements looks closest, if you know

Send inquiry

Opens your own mail client with my address and a short outline already filled in — nothing is submitted to or stored by this site. Prefer to write it yourself? marcin@rutkow.ski

Direct

marcin@rutkow.ski

Consulting, advisory and everything else. Typically respond within two business days.

LinkedIn

linkedin.com/in/rutkow

Background, roles and endorsements.

Heavy industry robotics

marcin@cayley.ai

Cayley AI business — heavy industry robotics deployments, full-team engagements and product work that belongs with the company rather than with me personally. More at cayley.ai.

Before you write

Please don't send anything export-controlled, classified or otherwise restricted over email. If your program is sensitive, say so in general terms and we'll set up an appropriate channel first.

Context

Twelve years of shipping the unglamorous half

Underwater vehicles navigating with no GPS or RF. Autonomy suites landing on production airframes. Industrial fleets supervised at a 1:N ratio. Cloud-controlled actuation on high-pressure wellheads. The pattern is always the same — the hard part is the seam between a capability and a deployment.

See the projects About me