
Physical Robotics builds intelligent humanoid robots that work safely alongside humans in real-world environments. The Pi robot uses force-controlled hands with high-bandwidth sensing, compliant actuators, model-based control, and data-driven policies to enable dexterity. It targets industries like manufacturing and healthcare and integrates with existing workflows and safety requirements. The platform emphasizes Physical Intelligence, transferring skills from simulation to reality. The company aims to scale deployment across logistics, field service, and other contact-rich tasks.

Physical Robotics builds intelligent humanoid robots that work safely alongside humans in real-world environments. The Pi robot uses force-controlled hands with high-bandwidth sensing, compliant actuators, model-based control, and data-driven policies to enable dexterity. It targets industries like manufacturing and healthcare and integrates with existing workflows and safety requirements. The platform emphasizes Physical Intelligence, transferring skills from simulation to reality. The company aims to scale deployment across logistics, field service, and other contact-rich tasks.
Physical Robotics | Norway (Oslo–Viken) | Core Team | Make History
We’re not hiring “an AI engineer.” We’re recruiting a championship AI builder
.
Humanoid robots will become the next industrial revolution: a new labor layer that can do real work in factories, logistics, and environments humans avoid. Most teams will publish papers or ship demos. A few will build production-grade physical intelligence
that survives the real world.
We’re building that team.
Small team. High standards. Big ambition.
If you want a career where you can point to a new chapter in humanoid AI and say “I built the brain that made it real”
The mission
Physical Robotics is a Norway-based humanoid robotics company founded by Dr. Phuong Nguyen
(ex–cofounder of 1X Technologies
) and backed by Skyfall Ventures
together with leading Norwegian industrial partners and investors. Our objective is to build a world-class humanoid
by unifying high-performance hardware
with physical intelligence
-the capability to perceive, plan, and act robustly in real-world environments.
We are building for the next inflection point: broad industrial deployment of humanoids in manufacturing and operations. We expect adoption to accelerate in 2027–2028
, and our strategy is to arrive at that wave with a platform that is production-grade, scalable, and operationally reliable
-not a lab prototype. We benchmark ourselves against the strongest global players- Tesla, Figure, Boston Dynamics, Toyota, Xiaomi, 1X Technologies
-and we intend Physical Robotics to become a reference name the field follows.
At Physical Robotics, we are building the AI foundation
for physical intelligence: Models that don’t just generate text-they perceive
, reason
, plan
, and act
under real-world constraints. This is AI that touches reality: Data quality, latency, safety, distribution shift, evaluation, deployment, and iteration speed.
You will help set the pace, the bar, and the architecture that the company scales on.
What you will own
“Make History” tasks you might lead
Build a training + evaluation pipeline that turns raw multimodal data into reliable policy behavior.
Required (must-have)
Preferred (high-leverage)
The mindset we’re recruiting
You are:
Obsessed with measurable progress
Why Physical Robotics
Core team impact
: your AI decisions define years of platform capability
Real deployment
: your models will run in production pipelines and on real robots
Competitive salary + incentive option program
A chance to build technology the world will copy and to make Physical Robotics a name the field follows
Apply
If you’re driven by hard problems and high standards, apply even if you’re not sure you qualify. If you can solve, learn, and ship, we want to talk.
Send your CV to giang@physicalrobotics.com
Subject: *AI Expert / Senior AI Engineer -Humanoid Robotics - Your Name
Core model strategy
: choosing the right architectures, objectives, and training recipes to win in real-world settings
Distributed training at scale
: multi-node training, performance tuning, memory optimization, reliability of long runs
Post-training + alignment
: fine-tuning, preference optimization / RLHF-style workflows, safety & behavior shaping
End-to-end AI pipeline
: data curation, labeling/filters, training infra, evaluation, deployment, monitoring
Serving & deployment
: low-latency inference, model packaging, versioning, rollout strategy, rollback safety
Real-world integration
: AI that operates inside a robotics stack (latency budgets, failure modes, sensor noise)
Ship a scalable distributed training stack that consistently produces strong models (not one-off runs).
Create post-training workflows that improve capability while keeping behavior stable and aligned.
Deliver a model serving system that meets real-world constraints: latency, uptime, monitoring, safety.
Push the frontier of Physical AI
for humanoid manipulation and autonomy.
Turn “research results” into deployable production artifacts
that run on real robots.
Deep expertise in at least three
areas:
Physical AI / Multimodal LLMs
Hands-on experience with distributed training
and multi-node training
Model training, optimization, serving, and deployment
Experiences in deployment techniques (model serving, designing data pipelines, AI pipelines).
Minimum four years of experience in A.I/ML engineering.
Proficiency in Python and related ML frameworks such as Torch, Tensorflow, and Jax.
Experiences in LLM post-training techniques ( LoRA/QLoRA, RLHF, alignment
).
Strong understanding of generative AI ( autoregressive model,
diffusion models, flow matching
).
Experiences in distributed training on large-scale datasets.
Strong background in ML system design,
and understanding MLOPs best practices.
Strong background in Mathematical and Machine Learning.
Master level degree in Computer Science, Mathematics or related field.
Strong publication record (top AI/ML/robotics conferences or journals). Preferred first author of publications in top tier A.I conferences such as Neurips, ICLR, ICML, AAAI, IJCAI, CVPR, ICCV, ACL, BMVC is a plus.
Kaggle experience (competitions, medals, high rankings).
Experience with Vision–Language–Action (VLA)
models.
Experience deploying models into real-world
or robotics systems.
Experience with DeepSpeed
/ Accelerate
is a plus.
PhD in Computer Science, Mathematics or related field.
Understanding of techniques such as test-time scaling
, MoE
, MPO
is a plus.
(not vibes): strong evals, ablations, reproducible wins
Comfortable owning the full loop: data → training → evaluation → deployment → monitoring → iteration
Pragmatic: you can do research-quality work and still ship production systems
Calm under pressure and relentless in debugging distributed training and deployment issues
Motivated by hard problems, not comfort
Ready to earn a seat on a small team doing world-class work
This role is for people who want to compete at the top level because humanoids are a global race.