Mechanical Engineer, Evals

What we do
At Arena Physica, we're building electromagnetic superintelligence. Our AI platform Atlas operationalizes physics-grounded intelligence to verify, debug, and optimize hardware across its lifecycle. Atlas is already trusted globally by the world's most advanced hardware companies, including AMD, Anduril, and Bausch & Lomb, for applications across R&D, integration testing, production assembly, and field repair.

About the role
As a Mechanical Engineer, you will advance Atlas's ability to reason about fundamental mechanics and build out our product features for real-world mechanical design and verification workflows based on structural, thermal and dynamic analysis. You'll work across a varied field of applications, from airframes, towers and avionics enclosures for aerospace and defense, to thermal management of high-power electronics, gimbals and mechanisms, and manufacturing equipment. Within this role, you'll work at the intersection of our hardware benchmarking team, which builds the evaluations, reward signals and training data used to measure, fine-tune and train models with reinforcement learning on real hardware engineering work, and our Engineering team, focused on operationalizing physics-grounded intelligence to verify, debug and optimize our customers' hardware across its lifecycle. You will do the work yourself in CAD, FEA and the lab, and turn it into the methodology that trains and informs our models, all in pursuit of redefining hardware innovation.

How you will contribute
Break mechanical workflows into into datasets for evals and RL/model training. Do real mechanical workflows yourself across the full hardware lifecycle, from design and simulation through build and test to failures in the field, using the tools a mechanical engineer actually uses.
Build the scoring and judging methodology. Define how LLMs are evaluated on mechanical work: what counts as correct, which criteria are checked automatically and which need an LLM judge, and how those criteria are weighted.
Automate "expert workflows." Convert best-practice mechanical analysis processes into automated pipelines, covering geometry cleanup and defeaturing, meshing strategies and convergence studies, boundary conditions and contact definitions, load-case setup, material models, and design-of-experiments sweeps.
Lead private evals with partners. Work directly with engineers at partner companies to pick the mechanical work worth measuring, turn their past designs and failure cases into evals.
Build mechanical capabilities for Atlas. Develop and maintain robust workflows for structural, thermal and dynamic analysis of parts, assemblies and systems, and make sure outputs are production-grade and repeatable.
Bring mechanical design expertise. Work in CAD and FEA tools such as SolidWorks, NX, Creo, Ansys and Abaqus, and support integrations and workflows into Atlas.
Support forward-deployed engineering. Embed with customers when needed to diagnose real hardware issues.
Represent Arena. Be a technical voice for Arena with partner engineering teams, researchers and the wider mechanical engineering community.
Travel domestically and internationally (up to 25% of your time).
Work in person at Arena Physica's NYC headquarters when not traveling.

Qualifications
5+ years of professional mechanical design experience in aerospace, defense, robotics, consumer electronics, automotive, energy or a related field, including taking designs through to production hardware.
CAD proficiency in at least one major tool (SolidWorks, NX, Creo, CATIA, Onshape or Fusion). Experience with scriptable or open-source CAD (FreeCAD, CadQuery, build123d) is a strong plus.
Analysis experience with FEA for structural and thermal problems (Ansys, Abaqus, Nastran, COMSOL or CalculiX), plus the ability to check results with hand calcs and know when a simulation is wrong.
Design-for-production knowledge, covering GD&T, tolerance stack-ups, DFM/DFA, material selection, fasteners and joints, and common manufacturing processes (machining, sheet metal, injection molding, additive).
Python proficiency. You can script analysis checks, parse CAD and simulation outputs, and work alongside software engineers.
Clear technical writing. You can explain a design decision precisely and write criteria that another engineer can understand clearly.
Bonus: Exposure to AI and experience creating training data, reward signals or grading rubrics for model training, reinforcement learning or LLM benchmarking, especially for engineering tasks.