Co-Founder · Cayley AI
I'm Marcin Rutkowski. For over a decade I've led the engineering behind machines that have to work unattended — underwater, in the air, on industrial sites and in contested environments. Today I build the deployment stack that gets physical AI out of the lab and into the field at fleet scale.
Technical leadership
Autonomy programs rarely fail at the algorithm. They fail at the seams — the link that drops, the firmware that can't be updated, the fleet that can't be operated by fewer people than it has machines. I work on the seams.
End-to-end autonomy stacks for vehicles that operate without supervision, and the system architecture that makes them auditable rather than magical.
Standardizing what runs on the machine — compute modules, sensor suites and firmware — so AI models can be deployed across a fleet instead of one prototype.
Ultra-low-latency, high-bandwidth links for supervised autonomy — and control planes that keep working as the link degrades instead of falling over.
Getting hardware to the site and keeping it running. Health management, readiness automation, and the bridge between engineering, deployment and the customer.
Co-founder · Present
Co-founded to build the physical AI deployment stack required to scale autonomous robotics across heavy industry and defense.
The models are arriving faster than the infrastructure that can field them. Cayley works on the layer underneath: telemetry and control that stay reliable in unstructured environments, edge orchestration, standardized embedded hardware, and sensor suites integrated for multi-asset coordination — aimed at the places where conventional solutions either fail outright or become cost-prohibitive.
Selected work
The deployment stack being built now, supervised autonomy across a heavy vehicle fleet, and an autonomy suite integrated onto a production airframe.
Flagship Cayley AI
The infrastructure that gets autonomous robotics fielded across heavy industry and defense — telemetry, edge orchestration and vehicle autonomy.
HavocAI (Teleo)
Ultra-low-latency comms and video pipelines for heavy vehicle fleets, letting one operator supervise many machines across an industrial site.
Shield AI
Led integration of the Hivemind autonomy suite onto the V-BAT platform, re-architected its avionics and set the organization's edge-compute standard.
Trajectory
Every step moved closer to the same problem: machines that have to make their own decisions, in places where help isn't coming.
Present
Co-Founder
Cayley AI
Leading engineering for the physical AI deployment stack — resilient infrastructure, vehicle autonomy, edge orchestration and sensor integration for multi-asset coordination across heavy industry and defense.
May 2024 — Sep 2025
Principal Engineer
HavocAI (Teleo) · Palo Alto, CA
Architected ultra-low-latency, high-bandwidth communication infrastructure for supervised autonomy over heavy vehicle fleets on large industrial sites — unlocking a 1:N operator-to-machine ratio, and bridging sales, deployments and operations to the infrastructure that had to support them.
Dec 2022 — May 2024
Senior Staff Software Engineer
Shield AI · Dallas, TX
Integrated the Hivemind autonomy suite onto the V-BAT, re-architected legacy avionics onto a distributed microcontroller platform, standardized the NVIDIA Jetson edge-compute ecosystem, and built the vehicle health and precision-landing systems behind a 100% readiness envelope.
Nov 2020 — Dec 2022
Robotics Software Engineer
Tankbots (Advanced Robotics Group) · Houston, TX
Architected a first-of-its-kind AUV for 24-hour endurance missions in zero-connectivity environments, including a proprietary underwater visual odometry system for navigation with no GPS or RF.
Dec 2019 — Nov 2020
Software Engineer
Microsoft (The Marsden Group) · Greater Houston
Embedded AI and computer vision on Jetson platforms — 4× faster CNN inference via Tensor Cores, a 500× improvement in LiDAR pipeline execution time, a 20× network latency reduction, and end-to-end hardware/software integration.
Dec 2016 — Dec 2019
Software Engineer
Schlumberger · SLIIC, Sugar Land, TX
Owned ValveCommander end to end — inception, architecture, engineering, manufacturing, field trials and commercialization. Also built characterization systems for nuclear detector photomultipliers and radiation detector accuracy validation.
May 2014 — Dec 2015
Embedded Software Engineer → Forward Deployed Engineer
Schlumberger · Sugar Land, TX & Al-Khobar, Saudi Arabia
Built a fiber-optic acquisition device capturing high-frequency data from downhole perforations through FPGAs and an ARM MCU over TCP/IP — then went to the field, running rigless well intervention with a team of five in the Saudi desert.
Consulting & advisory
That gap is where most autonomy programs stall. I take on a small number of engagements each year — architectural reviews, autonomy development guidance, and specialized subsystem and integration work.