Staff Machine Learning Engineer
Posted Updated
Description
- Sonatus is a global leader in the automotive industry, providing key technologies that enable intelligent AI-defined vehicles
- Our solutions are already on the road with millions of vehicles, and we are quickly expanding our offerings for production-grade AI on the Edge
- We are looking for a great Staff AI Engineer to join our seasoned AI team and lead the development of Edge AI for in-vehicle self-aware health monitoring and prediction
- In this role, you will build and deploy AI models that analyze continuous data generated in the vehicle during the day-to-day operation, including system logs, traces, and vehicle internal signals (Ethernet and CAN) to detect and predict the health of different sub-systems and anticipate failures in real-time
- You will own the end-to-end ML pipeline—from data ingestion and model training to deployment on resource-constrained edge devices and model optimization
- You will work in a fast-paced startup environment where your code will directly impact fleet reliability and build the next generation of the self-aware vehicle
- You will be expected to collaborate with other leading developers who have a deep understanding and expertise of vehicle software and systems, and other AI developers working on MLOps and integration of AI models on vehicles expected to be on the road today
- Expect to experiment with cutting-edge model architectures and best-in-class development tools
- This is a hybrid role out of our Sunnyvale, CA, where you will be expected to work in our office 3 days a week
- Build and train AI Edge models (e.g., Transformers, LLM, CNN, LSTM, Trees) to process unstructured application logs, kernel traces, and multi-modalities
- Integrating ML flows, including cloud-based LLM APIs (Gemini, OpenAI, Claude), with emphasis on synthetic data creation
- Develop algorithms to automatically cluster log patterns and detect software regressions, race conditions, or crash precursors
- Design unsupervised and supervised learning models (e.g., Autoencoders, Isolation Forests) to monitor time-series data from CAN bus and on-board sensors
- Implement logic to correlate signal anomalies (e.g., ADAS drifts, sensor spikes, latency jitters) across different modalities with system events to identify root causes
- Port and optimize PyTorch/TensorFlow models into production-grade models for execution on CPU/GPU-bound targets or embedded NPUs
- Apply quantization, pruning, distillation, and memory optimization to ensure models run within strict RAM/Flash budgets
- Define the data strategy for on-device filtering: pre-processing on device and decide which data is processed locally versus processed in the cloud
- Lead the architecture for the edge ML pipeline and mentor junior engineers on best practices for embedded AI
Benefits
- Stock option program
- Flexible working arrangements
- Unlimited PTO
- Comprehensive benefits with generous allowance
- Free in-office food & snacks!
- 7+ years in Machine Learning Engineering, with 3+ years focused on Edge AI or Embedded Systems
- Experience with NLP techniques for textual data parsing, sequence modeling (RNN/GRU), vector store, or lightweight LLMs/SLMs
- Proven ability to lead technical projects from concept to production in an ambiguous, fast-paced environment. Ability to communicate with stakeholders and articulate trade-offs
- Experience deploying to Edge environments (e.g., ARM-based), managing memory manually, and working with limited compute resources
- Proven experience mentoring junior engineers in software development
- Bachelor’s degree in Computer Science, Electrical Engineering, Software Engineering, or a related field
- Expert Python (for training) and decent working knowledge of modern C++ (C++14/17 for inference)
- Experience with libraries like scikit-learn, tslearn, or statsmodels for anomaly detection on sensor data
- Candidates with a strong Computer Vision (CV) / ADAS track record are highly encouraged to apply!
- Deep proficiency with PyTorch or TensorFlow, and experience with inference engines like ONNX, TFLite, or TVM
- MS/PhD in Computer Science, Engineering, or related fields
- Understanding of Linux/QNX kernel logs (dmesg), process states, and OS-level debugging
- Familiarity with Edge systems and preferably automotive formats (CAN, DBC, UDS, SOME/IP, or MQTT
- Experience with NVIDIA TensorRT, Qualcomm SNPE
Skills
Sign in to read and join the discussion.