The Team

FROM MODEL TO MACHINE.

We build AI-native, driving top-tier models across the whole stack the way we do at Timerise, so we design, ship, and scale enterprise-grade software - with a crew sized to start the first stage. Hardware is not new to us either - we built an AR shooting range that pairs live fire with projected targets and automatic hit detection, and MagicVR, a VR venue running our own games on rigs we built ourselves.

Skip to the challenge ↓

02 / The Crew

AI-NATIVE. ENTERPRISE OUTPUT.

One crew owns the whole stack today - strategy, software, design, and story.

Mateusz Sowa

Senior Engineer

Owns strategy and the entire software stack - from simulation infrastructure to the balance controllers that keep BASILISK on its feet. Two decades shipping backends, CTO at Timerise, founder before that. On the hardware side: built MagicVR's weapon controllers on airsoft frames wired straight into the game, and designed and coded the thermal-vision-powered AR Shooting Range. Shot detection in milliseconds - proof of the precision we need in software this fast.

LinkedIn ↗

Piotr Kozak

Product & Design

Owns how everything looks, feels, and reads - product, brand, and the walkout. A decade of product and UX design (LiveChat, Opera, intive) turned toward hardware, plus the program work: partners, league relations, and the calendar that ends in a fight.

LinkedIn ↗

Maciej Pelczar

Communications

Owns everything that happens outside the workshop - press, media, and business relationships. Runs contact with partners and sponsors, sets up commercial conversations, and makes sure the people who can fund or buy the next stage hear about it first. Carries the story out, turning a garage project into a league fixture.

hello@azimuthpro.com ↗

04Open seat

Engineer

Open

Full-stack builder who moves fast with top-tier models and cares how the result holds up. Grows into the discipline that fits - controls, perception, simulation, or ops.

Win this seat

03 / The Challenge

JOIN THE GRID.

A seat on this crew is not filled - it is won. Pitch the edge you would give the machine and back it up.

The prize

The adventure of a lifetime

Months inside a real fight program - building, breaking, and rebuilding a machine that has to answer in milliseconds.

Shenzhen, Season 2

Fly with the crew to China for the league. Fight nights, the pit, the walkout - from the inside, not from the stream.

The permanent dream job

Hold your ground through the season and the seat stays yours - a permanent role on the stack you helped build.

How it works

01Pick the area - or areas - where you fight best.
02Send your pitch: an idea for an advantage, not a cover letter.
03Every application gets an answer. Every single one.
04A maximum of 20 candidates go through to interviews.

The areas

Four disciplines, one machine. These are not job listings - they are the grounds the challenge is fought on. Pick where you are strongest.

01
Embodied AI / RL Engineer
Simulation
Builds the robot's brain. Designs and trains reinforcement-learning policies in simulation, owns the sim-to-real transfer, and optimizes the networks for concrete fight tasks - from manipulation to landing strikes in the URKL ring.
Python · PyTorch · Isaac Gym · RL / PPOClaim this area
02
Motion Control Engineer
Controls
Owns the physics of movement. Combines rigid-body dynamics and whole-body control with learned models so the machine keeps its balance when pushed, stands firm on uneven ground, and moves fluidly without damaging its own structure.
C++ · Control theory · Inverse kinematics · WBC (MIT Cheetah-class)Claim this area
03
Embedded / Systems Engineer
Operations
The bridge between code and hardware. Configures the real-time OS on the robot's on-board computer, optimizes the build system, talks to the actuators over low-level buses, and monitors battery and cooling under fight load.
C++ · Linux RT · ROS 2 · CMake · CAN / EtherCATClaim this area
04
Perception / Computer Vision Engineer
Perception
Makes the machine see. Integrates and processes the 360° LiDAR and depth cameras, owns SLAM and terrain mapping, obstacle detection, and locating the opponent in real time.
C++ · ROS 2 · OpenCV · PCL · YOLO-class detectorsClaim this area

04 / Apply

CONVINCE US.

One form, no cover-letter theatre. Tell us who you are, pick your ground, and pitch the advantage you would give the machine.

DON'T SEE YOUR GROUND?

The challenge is not limited to the four areas. If you can make the machine faster, steadier, or smarter - mark your area as Other and convince us.

Pitch anyway Become a partner