Patrick Rossi
Simulation and Infrastructure Engineer for Automotive Industry
A seasoned software and simulation engineer driving the design, deployment, and scaling of end-to-end simulation pipelines and infrastructure for autonomous vehicle development. With a Master's in Robotics and deep expertise in both simulation software and cloud-native compute environments, Patrick has over five years of experience architecting robust, high-performance frameworks—from launch-to-land flight sims at SpaceX to real-time autonomy simulations at a leading EV OEM.
Goals
Architect and optimize large-scale simulation pipelines that support full vehicle autonomy validation
Build and maintain infrastructure for HPC-grade simulations with automated configuration management
Integrate multimodal perception models into distributed simulation environments with sub-millisecond sync
Enable rapid iteration of scenario-based testing, including rare edge-case and hardware-in-the-loop scenarios
Leverage infrastructure-as-code and CI/CD practices to ensure reproducible, auditable simulation runs
Technical Skillset
Simulation Engines & Frameworks
CARLA
NVIDIA DRIVE Sim
LGSVL
Custom C++/Python engines
Infrastructure & Orchestration
Kubernetes
Terraform
Helm
Slurm
AWS Batch
GCP Managed Instance Groups
AI & Perception
PyTorch
TensorRT
OpenCV
ROS 2
mmDetection
TorchScript
Data Pipelines
Apache Kafka
RabbitMQ
Apache Arrow
REST/gRPC services
Sensor stream multiplexing
Embedded & Edge Deployment
NVIDIA Jetson AGX Orin
Drive PX2
QNX
AUTOSAR Adaptive
Real-time Linux kernels
Environment & Tools
| Domain | Tools & Technologies |
|---|---|
| Compute Infrastructure | Kubernetes clusters, Slurm queues, GCP Deep Learning VMs, AWS HPC instances |
| CI/CD & IaC | Jenkins/GitHub Actions, Terraform, Packer, Helm, Docker, GitOps workflows |
| Simulation & Testing | CARLA, NVIDIA Omniverse, Ansys AVxcelerate, custom scenario generators |
| Sensors & Data | LIDAR, RADAR, RGB/IR cameras, IMUs, CAN bus telemetry, sensor-fusion middleware |
| Monitoring & Logging | Prometheus, Grafana, ELK Stack, OpenX |
Pain Points
- Managing thousands of daily simulation runs with minimal human intervention
- Ensuring consistent, reproducible simulation outcomes amid frequent model and scenario updates
- Scaling GPU-accelerated sims while keeping storage and network I/O performant
- Automating validation of legacy configurations across evolving sim pipelines
- Generating and curating edge-case scenarios that accurately reflect rare but critical driving conditions
Metrics of Success
Pipeline Throughput
Number of full-vehicle sim runs completed per day; reduction in pipeline update time (target <30 min)
Resource Efficiency
Average GPU-hour per scenario; build-to-run ratio for containerized sim jobs
Coverage & Robustness
Percentage of edge-case scenarios simulated; false-negative rate in safety-critical tests
Deployment Stability
CI/CD failure rate; mean time to recovery (MTTR) for infrastructure outages
Configuration Compliance
Automated drift detection rate for sim configs; reduction in manual patching errors
Example Use Cases
Full Launch-to-Land Vehicle Simulation
Overhaul of a legacy pipeline to support end-to-end sim of launch, cruise, descent, and parking maneuvers—automating 500+ configurations.
Autonomy Scenario Orchestration
Defining complex, multi-agent test scenarios (e.g., urban intersections with pedestrians and cyclists) and scaling them across GPU clusters via Kubernetes Jobs.
Hardware-in-the-Loop (HiL) Testing
Integrating real ECU boards into the sim loop to validate firmware under virtual traffic and weather conditions, ensuring ISO 26262 traceability.
Continuous Simulation Validation
Implementing GitOps-driven sim scenarios that automatically trigger nightly regression sims, collecting metrics and visual dashboards for engineering review.