Deepgram - Defense / Edge Tech Lead
Upload My Resume
Drop here or click to browse · Tap to choose · PDF, DOCX, DOC, RTF, TXT
Requirements
• 5+ years of experience in systems engineering, embedded computing, or edge AI deployment, with a track record of delivering production systems on constrained hardware. • Strong proficiency in C, C++, and/or Rust, with experience writing performance-critical code for resource-constrained environments. • Hands-on experience with model optimization for edge deployment, including quantization, pruning, knowledge distillation, or architecture-specific compilation. • Familiarity with edge inference runtimes such as ONNX Runtime, TensorRT, TFLite, or vendor-specific SDKs (Qualcomm SNPE/QNN, MediaTek NeuroPilot, etc.). • Experience with security-conscious development practices, including secure boot, encrypted storage, code signing, and secure deployment pipelines. • Strong understanding of hardware-software interaction — CPU/GPU/NPU architectures, memory hierarchies, power management, and how they affect model inference performance. • Excellent communication skills — you will be the technical face of Deepgram to hardware partners and defense customers, and you need to be credible and clear in both contexts. • It Would Be Great if You Had • Prior experience working on or alongside classified defense programs — you understand SCIFs, accreditation processes, and the operational constraints of secure environments, even if you do not currently hold an active clearance. • Experience with ML model optimization techniques at depth — custom quantization schemes, mixed-precision inference, neural architecture search for edge targets. • Familiarity with ONNX, TensorRT, or similar model compilation and optimization toolchains and their tradeoffs across hardware targets. • Defense or govtech industry experience, including familiarity with procurement processes, ITAR, FedRAMP, or DoD software development standards. • Experience with real-time audio processing on embedded platforms — DSP pipelines, audio codec optimization, or streaming inference on microcontrollers or edge SoCs. • Background in hardware evaluation and benchmarking — systematically comparing accelerators, SoCs, or GPUs for specific workload profiles.
Responsibilities
• Lead the technical strategy for edge deployment of Deepgram's STT and TTS models, defining the architecture for on-device, on-premises, and air-gapped inference across diverse hardware targets. • Optimize models for edge and embedded platforms, driving quantization, pruning, distillation, and runtime optimization to meet strict latency, memory, and power constraints. • Partner with Qualcomm, Motorola, and other hardware vendors to ensure Deepgram models run efficiently on their chipsets, collaborating on SDK integration, performance benchmarking, and joint go-to-market. • Support defense customer requirements through AWS NatSec partnerships, translating mission requirements into engineering deliverables and ensuring Deepgram's solutions meet the unique demands of government environments. • Design and build edge runtime infrastructure, including model packaging, deployment pipelines, OTA update mechanisms, and telemetry for devices operating in low-connectivity or disconnected environments. • Harden deployments for security-sensitive environments, implementing secure boot chains, encrypted model storage, tamper detection, and audit logging appropriate for defense and government use cases. • Benchmark and validate performance across target hardware platforms, establishing repeatable test suites for latency, accuracy, power consumption, and resource utilization. • Collaborate with Research and Engine teams to influence model architectures toward edge-friendly designs from the start, reducing the optimization burden at deployment time. • Provide technical leadership to cross-functional teams working on defense and edge projects, setting engineering standards, reviewing designs, and mentoring engineers on systems and optimization practices. • You'll Love This Role If You • You find deep satisfaction in making a 300M-parameter model run on hardware with 4GB of RAM — and still hit accuracy targets. • You want to work at the intersection of AI and hardware, where optimization is not optional but existential. • You are energized by partnerships with hardware companies and enjoy the back-and-forth of getting a model to sing on a new chipset. • You understand the unique dynamics of defense and government customers and can navigate their requirements without losing engineering velocity. • You believe that edge AI is the next major deployment frontier, and you want to define how speech AI gets there. • You prefer working on hard, constrained problems over open-ended research — you want to ship, not just publish.
Benefits
• Holistic health • Annual wellness stipend • Mental health support • Life, STD, LTD Income Insurance Plans • Work/life blend • Generous paid parental leave • Flexible schedule • 12 Paid US company holidays • Quarterly personal productivity stipend • One-time stipend for home office upgrades • 401(k) plan with company match • Tax Savings Programs • Continuous learning • Learning / Education stipend • Participation in talks and conferences • Employee Resource Groups • AI enablement workshops / sessions • For candidates outside of the US, we use an Employer of Record model in many countries, which means benefits are administered locally and governed by country-specific regulations. Because of this, benefits will differ by region — in some cases international employees receive benefits US employees do not, and vice versa. As we scale, we will continue to evaluate where we can create more alignment, but a 1:1 global benefits structure is not always legally or operationally possible. • Backed by prominent investors including Y Combinator, Madrona, Tiger Global, Wing VC and NVIDIA, Deepgram has raised over $215M in total funding. If you're looking to work on cutting-edge technology and make a significant impact in the AI industry, we'd love to hear from you!
No credit card. Takes 10 seconds.