Website:
frntl.ai
Job details:
What we do at FRNTL?
At FRNTL, we’re building advanced humanoids that are designed to operate in hostile, real-world environments; on Earth today, and in Space tomorrow!
We operate with small, high-ownership teams, owning a critical part of the humanoid platform end-to-end, as part of the same nervous system.
What are we looking for from this role?
You will own the Foundational AI stack that powers stereo perception, reasoning, and decision-making, and whole-body control for our humanoid systems.
- This is a deep research + applied engineering role. You will design novel model architectures, build massive datasets, train large-scale models, and push them all the way into deployment on real humanoid systems with tight latency, power, and reliability constraints.
- This role is intentionally scoped for a one-man army: someone who can think like a researcher, build like an engineer, and ship like a systems architect.
If you know what it takes to train large foundational models end-to-end and felt frustrated by organizational bottlenecks, this role removes them from Day 0!
Responsibilities:
- Design and develop custom deep learning architectures for robotics and embodied intelligence
- Own the full AI lifecycle: dataset creation, annotation strategy, data modeling, training, validation, deployment on edge-platforms
- Train million–billion parameter models using HPC clusters and NVIDIA CUDA-based GPUs
- Optimize models for real-world deployment via quantization, compression, and acceleration
- Build and maintain end-to-end MLOps pipelines for continuous training, evaluation, and improvement
- Collaborate closely with Pranav Durai, robotics software, controls, and hardware teams to integrate models into real systems
- Push models from research prototypes to production-grade, field-deployed systems
- Drive internal research direction and technical strategy for Foundational AI at FRNTL
Minimum Requirements:
- PhD in Artificial Intelligence, Machine Learning, Computer Vision, or a closely related field
- 6+ years of hands-on experience designing and training custom deep learning model architectures
- Proven experience with large-scale foundational model training (million to billion parameters)
- Strong background in dataset engineering: collection, annotation strategies, curation, and data modeling
- Extensive experience training models on HPC systems using NVIDIA CUDA GPU clusters
- Strong proficiency in Python, C++, and CUDA C++
- Hands-on experience with model quantization and optimization (ONNX, FP to INT, TensorRT, etc.)
- Experience building and operating MLOps pipelines for automated training, evaluation, and deployment
- Strong mathematical foundation: linear algebra, probability, optimization, and deep learning theory
- Ability to translate cutting-edge research into deployable, reliable systems
Strong Preferences:
- Publications in top-tier conferences or journals (NeurIPS, CVPR, ICCV, ICLR, ICML, etc.)
- Graduated from Ivy or Tier-1 institutions (CMU, Stanford, MIT, Berkeley, ETH, IIT, IISc, etc.)
- Prior experience working on embodied AI, robotics, or real-time perception systems
- Exposure to deploying models on edge or resource-constrained platforms
- Background in multi-modal architectures (vision, language, proprioception, action)
- Experience working in high-ownership, zero-silo research environments
Perks:
- Competitive compensation with meaningful equity (Base + ESOPs)
- Comprehensive medical insurance, because health is non-negotiable
Click on Apply to know more.