New opportunity

AI Systems, Model Optimization

Unconventional, Inc. · Mountain View, United States

About this role

As a Member of Technical Staff in AI Systems, Model Optimization, develop the training methods, optimization strategies, and infrastructure needed to run AI models efficiently on Unconventional’s novel, physics-based compute substrates. The role connects model architecture to physical silicon and tapeout. The posting lists Mountain View, CA and US Remote as locations.

Responsibilities

  • Build performance models and energy benchmarks that assess compute, memory, and energy trade-offs and track Pareto-optimal model and hardware configurations.
  • Partition and map complex AI models to hardware. Develop quantization, sparsity, pruning, and distillation approaches suited to the substrates’ physical constraints.
  • Apply quantization-aware, noise-aware, and sparsity-focused training to address analog compute constraints, including memory footprint, connectivity, precision, and noise.
  • Develop and optimize GPU kernels with low-level tools such as CUDA, Triton, or CUTLASS; profile and debug training and inference bottlenecks in complex ML codebases.
  • Translate model requirements into specifications for hardware and infrastructure teams and document findings for tapeouts.

Minimum qualifications: An MS/PhD or equivalent research or project experience in a quantitative field such as AI/machine learning, computer science, physics, electrical engineering, or applied math. Requires practical knowledge of the modern AI/ML stack, optimized GPU execution, performance profiling, and the system implications of architectures such as Transformers, Mixture of Experts, and diffusion models. Deep PyTorch experience is required, including its internals, torch.compile, DDP, and FSDP.

Preferred qualifications

Experience with production training frameworks such as Megatron-LM or DeepSpeed and large-scale distributed training; interest in co-designing training systems around unconventional hardware physics; and research or practical work on advanced compression or approximation, including noise-aware or physics-constrained training.

Stated benefits include health benefits, 401(k) matching, unlimited PTO, and complimentary meals in the Palo Alto office.

Skills for this role

Performance modelingEnergy benchmarkingModel partitioningHardware-aware trainingQuantization-aware trainingNoise-aware trainingSparsificationPruningModel distillationCUDATritonCUTLASSGPU kernel developmentProfilingPyTorchtorch.compileDistributed Data ParallelFully Sharded Data ParallelTransformersMixture of ExpertsDiffusion modelsMegatron-LMDeepSpeedDistributed trainingQuantization

Your skill match

Checking your profile…

YOUR NEXT STEP

Get interview-ready for this role

A focused preparation guide, built around this job’s responsibilities and requirements.

✦ AI-generated guide
Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.

Loading this role’s preparation guide…

KEEP EXPLORING

Similar AI jobs

Related skills and specializations in United States. Matched to this role, not your profile.

Explore more jobs
Finding similar opportunities…