Edge ML Engineer: Port & Optimize Models for Edge

Aimlroles; Northern (KY); Hybrid; USD 150,000 - 210,000; Full time

Job summary

Deepgram is seeking an Applied ML Engineer for the Partner Platform Engineering team to port speech models to edge and non-NVIDIA hardware. You will adapt models, validate on real devices, and own the deployment pipeline.

You’ll work with embedded engineers, quantify latency, and optimize for constrained devices, shipping models to customers rather than publishing research, with a senior-to-staff level title depending on experience.

Qualifications

• Experience deploying ML models to edge or non-GPU hardware in production.
• Strong Python and PyTorch, with production-quality engineering habits.
• Experience with quantization and precision tradeoffs (INT8, FP16, mixed precision).
• Ability to modify models to fit platforms by swapping operators and adjusting architecture.

Responsibilities

• Port Deepgram speech models to non-NVIDIA and edge platforms.
• Own serving-side model decisions for edge targets: quantization and precision choices.
• Validate port performance on real hardware and catch regressions.
• Build deployment paths for edge targets: packaging and versioning.
• Collaborate with Embedded AI Engineers on custom kernels.
• Partner with platform vendors on runtimes and toolchains.
• Feed edge constraints back to Research and Impeller.
• Take on adjacent production concerns at the edge.

Skills

Python

PyTorch

Edge deployment

Quantization

Model porting

CUDA

Metal

NEON

ONNX Runtime

TFLite

ExecuTorch

OpenVINO

Qualcomm AI Engine

Vendor NPU SDK

Tools

ONNX Runtime

TFLite

ExecuTorch

OpenVINO

Qualcomm AI Engine

Vendor NPU SDK

Job description

Deepgram is seeking an Applied ML Engineer for the Partner Platform Engineering team to port speech models to edge and non-NVIDIA hardware. You will adapt models, validate on real devices, and own the deployment pipeline.

You’ll work with embedded engineers, quantify latency, and optimize for constrained devices, shipping models to customers rather than publishing research, with a senior-to-staff level title depending on experience.