Automation at Scale: Powering Charm’s Distributed Infrastructure

The Weight of a World That Hasn't Been Built Yet

There's a moment in every hard mission when you realize the map you brought doesn't match the terrain you're standing in.

For us at Charm Industrial, that moment came when we looked at what we were actually trying to do: deploy pyrolysis at a scale no one has ever attempted. Remote field sites for biomass and injection. Distributed equipment. Hundreds of different data sources. A growing, sprawling system that was getting more complex by the day. Our operators started out doing what so many other operators do: logging readings by hand, transcribing sensor data, managing checklists on paper binders and disconnected spreadsheets. It wasn't broken, but it was fragile, and it was a ceiling.

Handwritten biomass consumption logs (left) and char inventory recording (right)

Every new site meant more manual work. Every new pyrolyzer meant more people staring at gauges and typing numbers into fields. The data we needed to make smart decisions existed, but it was trapped inside manual data acquisition and analog processes that didn't talk to each other, couldn't scale, and would buckle under the weight of what we were planning to build.

The Unlock: A Different Kind of Infrastructure

As Charm first moved from development to production, we made a decision that changed everything: we would treat operational software as a first-class, mandatory investment. Not an afterthought, not a back-office function, but rather a core part of what we were building.

That meant rethinking how data is managed, how operators work and interact, and how a lean engineering team monitors a system growing in complexity every month. Our architectural choice was a deliberate one: a robust control system topped with a simple and intuitive user interface that seamlessly moves data into the cloud.

The first transformation was building modern SCADA (Supervisory Control and Data Acquisition) and IoT (Internet of Things) systems; this ecosystem incrementally began linking our physical equipment directly to our operational software. Pyrolyzer reactor temperatures, well injection pressures, product flow meter totals; we integrated thousands of telemetry streams into a centralized data repository in real time. No one is writing them down. No one is typing them in. The data just flows.

4th Generation Pyrolyzer Operating Dashboard

One of the most accelerative and valuable outcomes of this data layer was the digital transformation of our Measurement, Reporting, and Verification (MRV) activities. Reliable, validated data workflows eliminated uncertainty and manual transcription, helping to improve our Time-to-Verification by 7x and reducing costs (check out our blog post here for more!).

This changed something fundamental: adding a new pyrolyzer to the network now adds valuable data, not overwhelming work.

What started as an operational upgrade has also become an engineering accelerant. Our growing, always-on repository isn't just a log of what’s happened, it’s now the foundation for predictive modeling, anomaly detection, and process optimization that simply wouldn't be possible if a person still had to touch every data point first.

Solving the Last 20%: Honoring What Can't Be Automated

But, as anyone who works in an industrial, hard-tech setting knows, automation isn't everything. Our operators are the boots on the ground and leverage their eyes, ears, and spidey-senses in ways no sensor can replicate. The sound of a machine transitioning into hotflow, the reading on an analog gauge, the smell of an overheating motor. These things matter and are irreplaceable in real-world operations.

That's where our manufacturing execution system (MES) came in: we implemented a tool called Manufacturo for the processes that still needed human hands and human judgment (more here!). Take biomass screening as an example: what used to mean hand-writing batch weights on a paper slip and dropping it in a collection box to be transcribed into a spreadsheet by someone else, is now a single digital work order with instructions, checklists, data collection, and safety acknowledgements built in. Across five work order types, we've eliminated roughly 40 hours of manual data entry per week, and that number only grows as we add sites.

Operator performing routine machine walkdowns using Manufacturo

The goal wasn't to replace the operator, it was to free up more time dedicated to genuine observation and decision-making. When paired with the robust automation layer discussed earlier, the true scaling impact was that our same crew of operators can now run drastically more production systems simultaneously. This unlocked the cost curve on a tactical, shop floor level, paving the way for massive operational gains (learn more about how we have reduced cost here!). 

Building for What Comes Next

Now, the infrastructure we've built is about to pay another dividend.

Our 4th and 5th generation pyrolysis systems are rolling into production (check out the video here!) and operate at a higher throughput where batch-style processes simply won't keep up. The burden of steady, continuous biomass inputs and corresponding bio-oil + biochar outputs make manual data entry not just inefficient, but infeasible. You can't have an operator hand-logging feedstock scale weights or bio-oil production volumes when the machine doesn’t stop.

To plan for this, we're taking the aggregate digital toolbox of our SCADA network, Ignition platform, and custom Manufacturo workflows and extending these to automate new parts of our operations. Conveyor throughput, automated scale readings, feed rates, bio-oil flow totalizers: these flow directly into our systems as sensor data, with Ignition feeding collected data straight into Manufacturo's databases so every work order stays complete and traceable without a human in the middle. Now, with shopfloor data integrated directly into our business layer, our supply chain can react to material inventory changes more quickly by scheduling delivery of more biomass to feed the pyrolyzers or scheduling the pickup of oil to be sent to the injection sites.

Our network isn't a single plant. It's becoming a distributed system of interconnected nodes. Injection wells in remote locations will call for oil from pyrolyzers deployed across regionally co-located sites, all harmonizing in a dynamic theater of operations. 

As we look a bit farther ahead to the tenth injection site, the hundredth pyrolyzer and eventually, the millionth sensor, the decisions we're making right now about data architecture, operator tooling, and process standardization are a key foundation to let us scale without an avalanche of operational overhead.

Charm’s distributed network across diverse assets and workflows

We are fortunate to have top engineers from best-in-class hard tech, food & beverage, oil & gas, automotive, and medical companies, including SpaceX, Smuckers, Chevron, Honda, and Lilly. Despite the wide range of corporate missions, an essential theme always emerges: the time to solve architectural system challenges is before you scale, not after. 

We're by no means at the end of this journey, but we've crossed enough thresholds to know that the path is real, and that what we're building works. 

Interested in what we're building? We're always looking for amazing people to help us scale and would love to trade notes with others working hard on distributed hardtech!