Software Engineer, Edge AI

Rivian and Volkswagen Group Technologies•Palo Alto, CA

5d•$116,300 - $159,950

About The Position

Rivian and Volkswagen Group Technologies is a joint venture between two industry leaders with a clear vision for automotive’s next chapter. From operating systems to zonal controllers to cloud and connectivity solutions, we’re addressing the challenges of electric vehicles through technology that will set the standards for software-defined vehicles around the world. The Connectivity Edge AI team owns the software in the vehicle that moves and makes sense of vehicle data. Our work covers broadly these areas: On-device AI. Running voice detection, speech and other ML models directly on vehicle compute so features keep working when the network does not. Vehicle data collection - The upload and retrieval path for large payloads such as camera clips and ADAS recordings, including throughput, reliability, and cost over Wi-Fi and cellular. Connected features - Connected features like live camera streaming and mobile video downloads. Our software ships on Rivian R1 and R2, last-mile delivery vehicles, and Volkswagen Group platforms. Most of what we build has to work across several generations of hardware and more than one operating system. This is a hands-on individual contributor role for an engineer early in their career who wants real systems problems. You will write C++/Rust for services that run inside a car, under fixed CPU and memory budgets, latency targets a customer can feel, and no ability to redeploy in an afternoon. Where you start depends on your strengths and what the team needs at the time. You will work on one of the following areas as a starting project: the telemetry platform, the high-throughput upload path for ADAS and camera data, embedded inference for the voice assistant, and customer-facing features built on top of all three. We care far more about evidence that you build hard things than about years of experience or the particular stack you have used.

Requirements

2-4 years of software engineering experience, or a BS or MS in Computer Science, Computer Engineering, Electrical Engineering, or a related field with equivalent demonstrated building experience.
Demonstrated evidence of building complex software systems. This can come from anywhere: hackathons, project or robotics competitions, a university tech club or engineering team, a substantial internship, open-source contributions, or production work at a previous job. We want to hear about something you built, what was genuinely hard about it, and how you got it working.
Strong Linux and systems fundamentals: processes, threading and concurrency, memory, inter-process communication, and filesystems, plus the ability to debug a running system with tools like gdb, perf, and strace.
Proficiency in either C++ or Rust.
Ability to do methodical debugging on large interconnected systems.
Clear written communication in design docs, tickets, and code reviews.

Nice To Haves

Experience with inference runtimes such as ONNX Runtime, TensorRT, TFLite, llama.cpp, ExecuTorch, or a silicon vendor SDK, including quantization and deployment to constrained hardware.
Experience with audio or speech models: speech recognition, text to speech, wake-word detection, voice activity detection, or audio signal processing.
Android development experience, particularly AOSP, the audio stack, HALs, Binder, or JNI.
Embedded, automotive, robotics, or other resource-constrained real-time systems.
Networking and large-scale data transfer: throughput tuning, gRPC or similar IPC frameworks, and observability tooling.
Contributions to open-source systems or ML infrastructure projects.

Responsibilities

Design and build native services in C++/Rust that run on vehicle compute across Linux, Android, and QNX targets.
Work on the telemetry platform: signal collection, metrics derived on the vehicle rather than in the cloud, diagnostics, and the tooling engineering teams use to configure what a fleet reports.
Improve the vehicle data upload path, including throughput and reliability over Wi-Fi and cellular, event-triggered and on-demand retrieval, and the cost of moving data at fleet scale.
Deploy and optimize ML models on the vehicle's NPU, including speech recognition, synthesis, and wake-word detection, and hold them to per-platform latency and accuracy targets.
Instrument what you build so it can be debugged in the field, then use that data to find and fix problems on real vehicles.
Benchmark, soak test, and defend performance budgets so regressions are caught before they ship.
Write design docs, review code, and debug on bench hardware and in vehicles.