Case Studies
Real problems. Measurable results. Here's what happens when you stop tolerating manual processes and start automating them.
Eliminating a 30-Day Manual Process
The Problem
Ford Credit was spending millions on slow, manual vehicle audits. Auditors drove hundreds of miles to physically count cars at dealerships, discovering discrepancies weeks after the financial impact. In a rising interest rate environment, this lag was expensive.
What We Built
An IoT telemetry platform that automated verification for 450,000+ vehicles, saving around $5 million annually in operational costs and interest rate carry. What used to take weeks now happens in real-time.
How It Works
I led the architecture of a high-scale event-driven system using Kafka, Google Cloud Pub/Sub, and Azure IoT Hub. The system ingests 600,000+ vehicle events daily with state management to ensure data integrity in a distributed environment. Handles out-of-order events through Kafka partitioning by vehicle_id and stateful buffering.
Whether your team is 5 or 500, the pattern is the same — find the manual bottleneck, automate it, measure the savings.
Cutting 20% of Weekly Engineering Bandwidth
The Problem
Every morning started the same way. An engineer clicking through 80 Postman requests to trigger compliance checks. One by one, like a human cron job. This ate up 20% of our team's weekly bandwidth. When someone got sick for a week, compliance checks just... stopped. That's when we knew we had to automate.
What We Built
I transformed the architecture to fully event-driven, reducing daily compliance runs by 90% and freeing up 10 engineering hours every week. This let us onboard 7 new enterprise recordkeepers with zero added overhead while improving throughput and reducing latency by 35%.
How It Works
I implemented load-aware scheduling that monitors database capacity before kicking off new jobs via AWS SQS and ephemeral Fargate tasks. Instead of blindly running all checks daily, the system uses database diffing to detect meaningful changes and only publishes events when there's actually a delta. Smart resource utilization that scales.
Manual processes that 'only take 20% of someone's time' add up fast. Automate them, and suddenly your team can innovate instead of maintain.
Protecting America's Most Stolen Vehicle
The Problem
The Ford F-150 is America's best-selling truck. Also its most stolen. Owners needed real-time protection, but tracking vehicles raises serious privacy concerns. We had to balance theft recovery capability with strict data privacy compliance.
What We Built
A privacy-first telemetry system now deployed in thousands of 2024 F-150s, providing owners with cloud-backed theft recovery. When a theft is reported, the system automatically increases tracking frequency to help police recover the vehicle faster.
How It Works
I architected a Zero Trust ingestion layer using GCP Pub/Sub and Kafka where all telemetry is anonymized at entry, with PII encrypted in a secure vault accessible only via ephemeral tokens during active recovery. The adaptive recovery mode transitions the vehicle's modem to high-frequency reporting when theft is detected, balancing recovery speed with cellular costs and battery drain.
Security and privacy don't have to be trade-offs. With the right architecture, you can have both — and ship a product customers trust.
Have a bottleneck like these?
Tell me about the manual process that's draining your team's time. I'll respond within 24-48 hours with next steps.
Start a Conversation