FLUID Water Meter

I cofounded FLUID in 2015 with talented partners to build an IoT water monitoring platform for homeowners. Over three years, with support from investors who believed in our vision, I built everything from ultrasonic signal processing to cloud ML pipelines. We produced about 100 devices and learned invaluable lessons about bringing hardware to market. Residential water damage costs billions annually, but monitoring systems were expensive commercial products. We saw an opportunity for affordable clamp-on sensors—no plumber required.

We started with an off-the-shelf ultrasonic meter to learn time-of-flight measurement. Our Kickstarter campaign brought in early supporters who helped fund custom clamp-on sensors with injection-molded enclosures. As the technical lead, I built most of the platform while managing contractors for PCB design, FPGA programming, and mechanical work. We progressed from 3D-printed prototypes to production units with proper tooling, built a calibration lab for training data, and shipped working devices. While we ultimately didn't reach the scale needed for sustainable manufacturing economics, we're proud of the complete IoT platform we built and grateful for the opportunity to tackle this challenge.

Leak detection requires pattern recognition, not just flow measurement. We trained ML models on calibration lab data. PLC-automated valves created labeled datasets with known flow rates—ground truth for training and validation. Anomaly detection learned baseline usage per household, then flagged deviations. A running toilet shows sustained low flow. A pinhole leak creates constant baseline. Models distinguished these from legitimate use like filling bathtubs. Usage classification identified fixtures by signature. Showers: 2-3 GPM sustained. Toilets: sharp spike plus refill. Faucets: intermittent on-off. Irrigation: scheduled high flow. Cloud infrastructure ingested 1Hz telemetry, storing time-series and running batch ML jobs.

Hardware has fundamentally different economics than software. Injection molding tooling costs tens of thousands before making a single part. Kickstarter funded initial tooling, but bridging prototype to production requires massive capital. Economics work at volume—but reaching volume needs upfront capital. MOQs dominate decisions. PCB manufacturers want 1,000+ boards. Injection molding becomes viable around 5,000+ units. Component suppliers have MOQs. You're negotiating batch sizes against unit economics against cash flow. Managing contractors across mechanical, electrical, and firmware requires tight integration. An enclosure change affects PCB mounting. Firmware optimization might enable smaller batteries, changing mechanical design. We produced 100 devices—enough to validate technology, nowhere near sustainable economics.

Hardware startups face a fundamentally different challenge than software. While we succeeded technically—building working sensors, reliable firmware, and scalable cloud infrastructure—we learned that hardware economics require a different playbook. Tooling, inventory, assembly, testing, and regulatory compliance all demand significant capital before generating revenue. To reach the volumes where unit economics become sustainable, you need either substantial investment or enough pre-orders to fund manufacturing directly. Despite the support of investors who believed in our vision and customers who backed us on Kickstarter, we didn't reach the scale needed to bridge that gap. The classic challenge: you need volume for viable unit costs, but you need capital to reach volume. Our 100 devices validated the technology, but sustainable manufacturing required thousands. We're grateful to everyone who supported us on this journey—our investors who took a chance on the vision, our Kickstarter backers who believed early, and our team who worked incredibly hard to bring this to life. The experience taught us that in hardware, technical excellence is necessary but not sufficient.

FLUID taught me more than any project before or since. We built technology that worked, learned deeply across domains from ultrasonic physics to ML to manufacturing economics, and had the privilege of working with talented cofounders, supportive investors, and early believers who took a chance on our vision. The experience shaped my approach to production systems. Shipping matters more than perfection. Real-world conditions humble lab results. OTA infrastructure isn't optional for IoT. Business viability requires more than technical excellence. Three years, 100 devices, countless lessons. I'm grateful for the opportunity to have tackled this challenge and for everyone who made it possible.