Senior Embedded AI Engineer

Sonatus is headquartered in Sunnyvale, California with international locations in Ireland, France, Germany, Poland, Korea, China, India, Taiwan and Japan. This is a hybrid position at our Pangyo location. You will report to the Engineering Director.

About the Role

Sonatus is bringing production-grade AI to the vehicle edge with Fastlane™ Edge, the in-vehicle AI runtime that lets OEMs deploy, run, monitor and update AI models, virtual sensors and diagnostic logic on existing vehicle computers. Our Customer Success organization operates as Mission-Driven Squads that take full end-to-end ownership of a module, from implementation to verification to on-site support.

We are looking for an Embedded AI Engineer to join our Korea Customer Success team and put edge AI into our Korean OEM programs. You will bring up the Fastlane Edge runtime on customer hardware, onboard AI models authored by partners, by the customer and by Sonatus, help customer engineers author their own models with the Sonatus SDK, validate models on real vehicles, and tune Sonatus in-house virtual-sensor models on customer data using our MLOps toolchain. You will act as a technical contact for the customer on edge AI integration topics.

This is a hands-on role for an engineer who is comfortable in a Python training pipeline, inside a Linux container on an automotive ECU, and in a vehicle at a customer test site. Model development is part of the job, but always in service of a customer deliverable.

Duties and Responsibilities

1. Edge AI Runtime Integration on Customer Hardware

• Hardware Bring-Up: Bring up and maintain Fastlane Edge on customer targets — Sonatus gateway units (Fastlane Copilot) and customer control units — including CAN/CAN-FD bring-up, DBC/ARXML integration, sandbox and Python-runtime deployment, security-policy adjustments and cloud connectivity.
• SDK Integration: Integrate the runtime with customer platforms via the Sonatus Southbound SDK where direct bus access is not available.
• Runbook Ownership: Maintain the Korea integration runbooks and default configurations; work with the platform team to turn one-off demo fixes into mergeable, released code.

2. AI Model Onboarding and Validation

• Model Packaging: Validate, package and deploy AI model bundles from partners, from customer engineers and from Sonatus using the Sonatus ADK: input-signal mapping, sampling and output schema review, dependency checks, resource budgets and OTA deployment through the Sonatus Cloud.
• Concurrency Verification: Verify multi-model concurrency on a single ECU within agreed CPU/RAM envelopes; measure and document per-model resource profiles and telemetry policies.
• Native Porting & Optimization: Migrate Python-developed models to production-ready native (C++/ONNX Runtime) bundles, including model conversion, native integration and optimization, with automated numerical-parity validation against reference implementations.

3. Customer Enablement

• Training & Office Hours: Deliver SDK/ADK training and office hours in Korean for customer engineers authoring their own models; support model-contract (config.yml) authoring and signal-to-input mapping from customer network databases.
• Requirements Capture: Help capture customer AI workload requirements (use case, signals, rates, accuracy/latency targets, resource and data-sovereignty constraints) and document them for Sonatus product teams.
• Documentation as Code: Treat documentation as code: integration guides, troubleshooting FAQs and release notes in Korean and English.

4. Model Development and Tuning on Customer Data

• MLOps Toolchain: Run customer vehicle captures through the Sonatus MLOps toolchain: data ingest and alignment, recipe-driven training, model-family A/B comparison (gradient boosting, temporal convolutional and physics-informed models), ONNX export, quantization and on-target latency measurement.
• Virtual-Sensor Tuning: Tune Sonatus virtual-sensor and estimation models on customer drives and deliver them as validated bundles with customer-facing reports.
• Ground Truth & Datasets: Define ground-truth capture (reference sensors, co-logging) and dataset versioning; contribute fixes and hardening upstream to the toolchain.

5. On-Vehicle Validation, Demos and Customer Support

• Validation Drives: Execute on-vehicle validation and comparison drives; contribute to acceptance criteria and shadow-mode protocols with the customer and Sonatus QA; produce traceable test evidence and, where required, ASPICE v4 MLE artifacts.
• On-Site Support: Support major test events and demos on-site, troubleshooting integration issues directly in the vehicle or on the bench.
• Defect Triage: Act as a technical contact for edge AI integration issues; reproduce, isolate (bundle vs. runtime vs. cloud), and follow customer-found defects through the Sonatus release train.

6. AI-Augmented Workflow

• AI Tooling: Leverage AI tools such as Claude and GitHub Copilot to accelerate recipe authoring, test generation, log analysis and documentation, within customer data-governance constraints.

Qualifications and Experience

[Essential Background]

• Education: Bachelor’s degree in Computer Science, Electrical Engineering, Mechanical/Control Engineering or a related field; Master’s degree preferred.
• Experience: 5+ years of experience in machine learning engineering or embedded software, including 2+ years deploying ML models on edge or embedded Linux targets. Customer-facing integration, field application or solutions engineering experience is a strong plus.
• Languages & Systems: Strong Python; ability to read, build and debug modern C++ (C++14/17); solid Linux debugging skills (systemd, processes and logs, containers; exposure to cgroups, sandboxing, seccomp or shared-library/glibc issues is a plus).

[Core Competencies]

• Structured Device/Signal Data: Hands-on experience working with structured device or bus-level signal data — automotive (CAN/CAN-FD), industrial fieldbus, or similar — including signal mapping, sampling and time alignment. Familiarity with automotive-specific protocols (DBC/ARXML, UDS, SOME/IP, DLT) is a plus, not a prerequisite.
• ML Deployment: ML deployment experience with ONNX/ONNX Runtime and PyTorch or scikit-learn/XGBoost; exposure to quantization and latency/memory profiling on ARM (aarch64) targets.
• Applied Modeling: Applied modeling experience with time-series or vehicle-signal data (regression, estimation or anomaly detection) and the ability to evaluate models against agreed metrics.
• ML Pipelines & Data Tooling: Experience with reproducible ML pipelines and data tooling such as Parquet, Apache Arrow or DuckDB, experiment tracking and versioned datasets.
• Verification Mindset: Python test automation and evidence-based triage; willingness to split time between development and validation as the mission requires.
• Communication: Fluent Korean and professional English; clear technical writing and comfort supporting customer engineers and on-site test events.
• Collaboration: Results-driven, with strong problem-solving skills and the ability to collaborate effectively with customer and internal teams.

Preferred Knowledge and Experience

• Platform Experience: Hands-on experience with NXP S32 family / eIQ Auto, Qualcomm SA8xxx / SNPE-QNN, or comparable embedded compute platforms.
• Vehicle Dynamics: Vehicle dynamics, chassis or body control-domain knowledge; prior virtual-sensor or sensor-replacement work.
• OEM IT & Security Stacks: Familiarity with OEM IT and security stacks (Kubernetes, Harbor/Nexus, Kafka, SonarQube/Blackduck, security review processes) and data-sovereignty requirements.
• Standards Awareness: Awareness of ASPICE v4 (Machine Learning Engineering process group), ISO 26262 and ISO 21448 as applied to ML components.
• LLM-Assisted Engineering: Experience with LLM-assisted engineering workflows (agent tooling, MCP) and evaluation.
• Tooling & Process: Proficiency with Jira/Confluence, Git-based workflows, and Agile/Scrum in cross-functional squads.
  
  

Korea Office Benefits & Perks Offered:

• Point Reward Program for Customer Integration team
• Support 4 social insurance plans
• Family leave (Maternity, Paternity)
• Hybrid office work-arrangement
• Complimentary lunches, snacks and beverages during on-site working days
• Wellness benefit allowance
• Phone & Internet reimbursement
• Computer Accessory Allowance
• English Support class
• Annual health checkup
• Holiday Appreciation Gift