Jacob Ogle

• Designing and building cloud-native data pipelines and engineering infrastructure to support Ford's ADAS products on GCP, including Dataflow, Cloud Run, Vertex AI, and GKE.
• Full-stack development of web applications supporting a transparent, user-friendly ADAS data platform for internal product teams.
• Workflow automation and scaling using GCP tooling: Terraform for infrastructure provisioning, Dataflow for data processing, and Vertex AI for ML workloads.
• Developing containerized services and distributed workflows spanning all phases of the data and machine learning lifecycle in collaboration with algorithm and vehicle platform teams.
• System integration testing, production deployments, and cloud infrastructure management with a focus on observability and CI/CD best practices.
• Operationalizing prototype ML environments and pipelines developed by algorithmic teams; defining KPIs and integrating telemetry to track cloud service efficiency.

• Owned and operated end-to-end data infrastructure supporting large-scale multimodal clinical AI research, expanding an inherited ~500 TB dataset to near-petabyte scale through integration of open source datasets and rolling clinical data acquisitions.
• Built and maintained production DICOM ingestion and preprocessing pipelines using dcm2niix with a custom metadata parser, generating study-level configuration files across a large-scale clinical imaging corpus spanning MRI, CT, and pathology modalities.
• Deployed and operated a self-hosted LLM inference stack on HPC GPU nodes using vLLM across 8 NVIDIA L40 GPUs, running summarization, annotation, and classification workloads over the full clinical text corpus — pulling reports from an internal Apache Solr instance with metadata managed in SQL Server.
• Owned Slurm job design, scheduling, execution, and monitoring for large-scale data processing and model training workloads on HPC infrastructure, ensuring reproducibility, versioning, and data quality across research workflows.
• Maintained a modern internal data catalog using DuckLake backed by DuckDB and SQLite, alongside SQL Server relational databases integrating heterogeneous clinical imaging, pathology, and report-level data sources.
• Supported vision and vision-language model training pipelines with early development of a post-training evaluation harness for assessing model performance across multimodal clinical tasks.
• Contributing author on ongoing peer-reviewed research in multimodal clinical AI.

• Served as the primary data engineering and AI authority on the plant IT team, independently architecting and deploying data infrastructure across molding, painting, assembly, and inspection lines for a full-scale automotive fascia manufacturing facility.
• Designed and deployed real-time PLC telemetry ingestion services in Java 21 across molding and paint production lines, running as daemon processes streaming press and line data to a Dockerized RabbitMQ broker, with a downstream Python-based alerting system enabling automated detection and notification of parts requiring quality inspection.
• Architected and deployed a plant-wide internal RAG system using ChromaDB, LangChain, and Ollama, indexing the facility's entire NorWeb knowledge base — reducing document retrieval latency from ~5 minutes to under 10 seconds for the entire plant workforce.
• Built centralized Python REST APIs backed by SQL Server and Azure SQL, serving ~30 upper management stakeholders across production, HR, and accounting departments as their primary interface for daily operational data.
• Engineered scalable data ingestion and validation pipelines using Apache Airflow, integrating multi-source sensor telemetry and enterprise data across production floor systems.
• Contributed to a multi-plant computer vision PoC using OpenCV and YOLO for automated assembly inspection, detecting bolt attachment defects on fascia production lines in real time.
• Contributed to hybrid cloud/on-prem architecture leveraging Azure Data Factory, Azure Blob Storage, and Azure SQL with focus on observability, reliability, and platform scalability.

• Embedded within the XC team supporting development of the GM central control gateway module responsible for OTA software reflashing and update management.
• Built automated build and code signing tooling in Python implementing GM key structure requirements, enabling compliant OTA reflashing workflows for the gateway module and forming a critical part of the software delivery pipeline.
• Contributed directly to the production gateway module codebase, shipping features and fixes in compliance with MISRA C standards and GM software development guidelines through a structured code review process.
• Developed an internal JIRA ticket monitoring tool for upper management, providing real-time visibility into ticket time-to-completion metrics across the XC organization.
• Co-led the Bosch XC Python Development Club in Plymouth, running sessions on Matlab-to-Python migration and practical data tooling using Pandas, NumPy, and Matplotlib for engineering workflows.

• Supported the Vehicle Health Alerts (VHA) team in diagnosing field-reported vehicle diagnostic issues, including TPMS anomalies and related ECU-level faults surfaced through cross-functional JIRA tickets.
• Used GCP BigQuery to extract and analyze diagnostic snapshot data for specific vehicles, collaborating with team members to identify data patterns and contributing findings to development team reviews.
• Gained hands-on exposure to vehicle diagnostic data architecture, ECU feature sets, and cross-functional automotive software development workflows.

• Developed clinical research participant tracking databases and data management frameworks supporting multi-institutional rare disease research cohorts.
• Supported large-scale computational genomics pipelines under the Gabriella Miller Kids First federal research initiative, including WGS/WES sequencing projects.

Developing custom firmware for an ESP32-CAN-X2 dual CAN controller to read and decode OBD2 vehicle data for real-time race telemetry. Hardware includes a soldered OBD2 pigtail interface and a UART-connected display implementing a functional RPM shift light on simulated CAN data. Built using Rust HAL targeting bare metal embedded deployment.