Inside the Industrial Autonomy Stack: Mining, Warehouses, and Perception Hardware from Pittsburgh

Open on YouTube ↗
Overview

Recorded live in Pittsburgh with the Pittsburgh Robotics Network, this episode of Manufacturing Happy Hour asks what industrial autonomy looks like when it leaves the lab and goes into mines, factories, and warehouses. Host Chris Luecke brought together three executives who each work on a different layer of that stack. David Griffin is Chief Sales Officer at Seegrid, which builds autonomous mobile robots for material handling. Mike Smocer is CEO of Mine Vision Systems, which captures real-time 3D data underground. Brett Phillips is Chief Revenue Officer and General Counsel at Hellbender, which designs and manufactures the sensing and edge-AI hardware that autonomous products rely on.

23 min read

Luecke split the conversation into three parts: concrete examples of autonomy in use and how quickly it is changing, how each guest's background shapes their approach, and why Pittsburgh has become a base for this kind of work. A consistent theme ran through all three. The panelists see physical autonomy as a problem of respecting existing workflows, doing careful engineering, and joining advanced technology to hands-on manufacturing, and they see the technology itself as only one part of that.

3:08

Three Companies, Three Layers of Autonomy

Each guest described their company in plain terms before getting into specifics. Phillips said Hellbender specializes in on-edge AI hardware development and called it the business of "manufacturing the eyes and the ears of the autonomy stack." In his account, two things set it apart. The first is a staff of "gray beard" engineers who have been doing this work for a long time. The second is that Hellbender handles both engineering (design and prototyping) and manufacturing under one roof, which he claimed makes it one of the only companies outside China to do so. What Hellbender mainly sells, he said, is a shorter time to market.

Griffin said Seegrid, like many Pittsburgh robotics companies, traces its origins to Carnegie Mellon. A professor there developed vision-based navigation "literally over two decades ago," long before vision navigation became common. Seegrid now moves materials inside industrial facilities for manufacturers, warehouses, and logistics companies, mostly with tow tractors and automated lift trucks. Its customers include both major and mid-tier manufacturers.

Smocer described Mine Vision Systems as a mining technology company. He warned that if someone has researched mining technology online, "there's a 99% plus chance" that what they found has nothing to do with his company's work. Mine Vision Systems creates what he called "measurable digital records" of a mine as it advances and the ore body is extracted. The 3D images are built in real time, with real-time location in a GPS-denied environment. In his view, the data feeds real-time decisions that carry millions of dollars of impact for companies pursuing critical minerals and precious metals.

8:41

Mining: Replacing Colored Pencils and Tribal Knowledge

When asked for a concrete example, Smocer first placed his company within autonomy. Capturing images and measurements is the sensing layer. He described a progression: once you capture what was "previously uncapturable," people can make better decisions first, then computers, and eventually machines act on that information.

Smocer said the value of the data is not in doubt, because mining companies "have been trying to capture it for centuries." In his account, the current method, without his company's technology, is "literally colored pencils and paper." Highly trained geologists walk up to the mine face and sketch what they see. The decisions made there affect the investment that flowed into the company over the previous decade, all the downstream processing, and the viability of the mine itself. He described these geologists as very good at their work but "completely underserved by inadequate information." Too much depends on tribal knowledge, results are inconsistent, and predictability is low. The problem is getting worse because few new graduates have these skills, so experienced people are retiring without anyone to inherit what they know.

He explained the economics in simple terms. Advancing an underground mine is "literally like building a tunnel and you don't know where it's going for the most part." The goal is to extract the valuable material without breaking more rock than necessary. Excess broken rock is waste that costs millions of dollars a year. Breaking too little means spending extra time on processing.

His example came from a customer that recorded data after every blast and fed it into a semi-autonomous process to update its mine plans day to day. The mine plan pointed one way. By reviewing the maps built blast by blast, in 10-foot by 10-foot increments, the team saw an opportunity in a different direction that was not visible to the eye. According to Smocer, they extracted $900,000 of value within three months of adopting the technology. He added that results of this size are "not uncommon" among his customers.

12:02

Seegrid: Replacing Labor at Large Companies, Augmenting It at Mid-Tier Ones

Griffin said manufacturing is currently Seegrid's largest business segment. With very large manufacturers, which he described as multi-billion-dollar companies with dozens of facilities, Seegrid has deployed hundreds of robots and, he said, saved them tens of millions of dollars a year.

The trend he emphasized from the past two to three years is a strong push from mid-tier, more regional manufacturers. He drew a distinction here. At larger manufacturers, automation sometimes means replacing labor. At mid-tier companies it is mostly about augmenting labor, taking on tasks the company "simply can't find people to do" because it cannot hire and keep enough workers. He described regional manufacturers deploying dozens of robots that cover the work of several dozen employees a year. He said this is what lets them scale and keep growing.

13:18

Hellbender: Building Perception Systems at Scale

Phillips said that Hellbender does not "currently" sell end solutions, and he flagged that word as foreshadowing. Today the company designs, engineers, and manufactures components that go into autonomy products, as well as some complete products. He mentioned that Hellbender manufactures Mine Vision Systems' products.

His main example was a Fortune 200 company building its own autonomy model for its vertical. That customer hired Hellbender to engineer, design, and manufacture a high-fidelity data collection and perception system, which he said is known in the self-driving world as a "tiara." Hellbender makes its own PCBs and does full box-build assembly in-house. Phillips said this lets it bring products to market much faster, both in getting them working and in producing them at scale. For this customer, Hellbender expects to produce "thousands and thousands of units in months," which he estimated would have taken years five or ten years ago.

How Fast the Capabilities Have Changed: From Pallets to "Anything Forkable"

Luecke asked each panelist how much autonomy has changed in just a few years. Griffin used automated lift trucks as his example. Only several years ago, he said, autonomous lifting was largely limited to standard wooden pallets moved floor to floor or floor to conveyor. More recently, better perception systems, control systems, and AI and machine-learning algorithms have expanded that. He said Seegrid's trucks can now pick up almost any forkable container and place it on any surface or location that will accept it.

He said this changes the scope of automation in a facility. Where there were once a handful of candidate applications, there are now "dozens and sometimes hundreds," because the perception and control capabilities have improved so much.

From a Three-Year, Multi-Million-Dollar Project to Months and Under a Million

Phillips said technology is advancing rapidly and cited Moore's law, which he described as doubling "every single year." He said Hellbender has experienced that pace directly. Much of its executive team spent years on autonomous projects at CMU, specifically at NREC, and worked on what he called the first safety-certified autonomous floor scrubber. By his recollection, that project took about three years and "millions and millions of dollars."

Hellbender recently signed a new customer to develop a roughly analogous product with a smaller form factor. Phillips expects it to take months and less than a million dollars in engineering cost. He credited several factors: open-source technology, faster progress in sensors and models, and advances in material science and form factors. He said his team is "standing on the shoulders of giants" and that the trend is continuing.

Mining: From Point Solutions to Systems, Then to Native Applications

Smocer spoke specifically about mining and made two points. First, the industry is moving from automating individual vehicles or pieces of equipment toward a system-based approach. He called this the natural course of technology in general, from point solutions to workflows. The system might be your own IP or someone else's, and he said acquisitions in the industry tend to follow this pattern.

Second, he placed mining in a sequence of adoption. In his view, discrete manufacturing adopted digitalization and automation first, then process industries such as chemicals and materials, and mining came last. He expects mining to follow a similar path but to cover the distance much faster.

He said the first technologies to get attention in a newly adopting industry are usually ones that have already worked elsewhere. Moving them over still requires real application and engineering work, but "the evidence is in, it works." He recalled a presentation from about three years earlier by one of the largest mining companies in the world, a customer of his. It laid out a vision for autonomy and digitalization built around fleet optimization, fleet orchestration, and predictive maintenance. Apart from one mining-specific term, "load, haul, dump," he said it "could have applied to any number of industries."

In the second phase, which he sees starting now, companies build native applications from the ground up for mining because there is no analog to borrow from another industry. He said Mine Vision Systems is not the only company doing this and that the two phases are proceeding together. He sees the trend both in his own company's decision loop and in other parts of the mining production workflow.

A Lawyer in the Revenue Seat

For the background segment, Luecke started with Phillips's legal training, which he linked to the liability questions raised by driverless vehicles on Pittsburgh's hills. Phillips said the combination of a revenue role and a legal background is unusual and that its benefits go beyond autonomy. As a salesperson moves up, he said, deals increasingly close through contracts. He can make the business case to decision makers and then negotiate "very arbitrary and niche clauses" with their counsel while understanding how those clauses affect the deal.

He said autonomy adds its own issues, especially product liability, and called indemnification "always the big scary one." In his view, the legal background lets him think more holistically than a pure sales focus would.

22:58

Change Management as the Real Discipline

Luecke asked Smocer how he ended up in mining after roles at Ansys (following its 2008 acquisition of Ansoft) and at a company doing machine learning for chemicals and materials. Smocer declined to call it happenstance and accepted Luecke's suggestion of "Providence." He then explained the common thread. In every industry he has worked in, he introduced computer-based technology, mostly for product development, to organizations that had operated successfully for a long time without it. Consumer electronics and semiconductors were possible exceptions. The work, in his view, is transformation and change management.

He warned against disrupting workflows because a new technology "solved one little problem and broke 12." Doing that damages a vendor's reputation and also puts customers' careers at risk. At Ansys, the breadth of physics solutions let him tell customers the company would innovate faster than they were likely to adopt, which made it a safe bet. That supported multi-year partnerships built on the understanding that the customer would keep going after the first success. He said he has seen technologies fail because their makers "played their cards a little bit too close." They did not think about workflows, adjacent technologies, or the machines and people in those workflows, and they did not understand the domain.

He applied the same thinking to mining. The industry is behind, he said, because technology has not served it and the solutions did not exist, and slowness to adopt is not the reason. Miners have been "remarkably innovative in the absence of all this technology," and vendors need to respect that. They also need to show customers a vision beyond the first point solution and ask how the customer wants them to grow. He described these lessons from 30 years of work as the part he is proud of.

26:33

A Software Company That Drags Hardware Along

Luecke noted that Seegrid is Griffin's first job in manufacturing and robotics. Griffin said the thread in his career is software. He started as a software developer and has always worked for what he called "software first companies that actually drag a lot of hardware along for the ride."

He said Seegrid is often seen as a product company because its products weigh a ton and move multiple tons of material. He considers it a software company. In his description, the hardware is "essentially a near commodity at this point." The value is in navigation, safety, perception, and control, which are software capabilities running on that hardware. He pointed to Seegrid's headcount: 12 software engineers for every hardware engineer. The end product, as he put it, is a result expressed as ROI, which makes his current role very similar to his earlier ones. Phillips joked that a former software developer now working in sales made Griffin "a real unicorn."

28:28

Why Pittsburgh: Pragmatism and Blue-Collar Innovation

Smocer and Phillips have deep Pittsburgh roots. Griffin is from Virginia and spent most of his life in Atlanta. Phillips grew up moving around as a military child, graduated high school in the region, and married into a Pittsburgh family.

Smocer has lived in Pittsburgh all but six years of his life and studied at the University of Pittsburgh as an undergraduate and graduate student. He pointed out, with a joke, that Luecke had left Pitt off his list of anchor institutions. He said he chose the startup community years ago because he wanted to take part in how technology, much of it robotics and automation, was changing the city. When he graduated, he said, some sections of the city people avoided are now its hottest spots. Healthcare has always been there, the steel industry moved out, and "it's really technology that filled that gap." He described himself as "comfortable being uncomfortable." He spent three years as CRO for a Silicon Valley company that he still advises, and he called Silicon Valley "the standard in terms of the success stories." What he values in Pittsburgh is its pragmatism, a "get things done type of attitude that doesn't have a lot of fluff around it." He avoided the word "grittiness" as overused.

Griffin said he has spent a lot of time in the region over the past five years and named two things that stand out. First, although Seegrid is an innovative robotics company, it is "a blue-collar business." Its products operate in dirty factories, busy warehouses, and fast-paced logistics environments, and most of its field employees do blue-collar work. He said Pittsburgh people combine that grit with innovative thinking. Second, the region allows cross-pollination. Seegrid hires many people from the University of Pittsburgh and also from local industries such as medtech and manufacturing.

Phillips agreed and tied the blue-collar point to physical AI. He said any autonomy system has to interact with the physical world, so "you have to have people that are willing to get their hands dirty." He said he would "die on the hill" that Pittsburgh's engineering talent matches Silicon Valley's, and argued that, given the history of autonomy stemming from CMU with Pitt also playing a large role, Pittsburgh could be called "the OGs" of the field. What he thinks will set the city apart going forward is the ability to combine high-level technical engineering with real-world manufacturing, including "burning your hands on soldering irons." He called that part of Pittsburgh's culture.

36:03

The Pittsburgh Robotics Network as a Business Engine

Luecke said PRN is one of the more active industrial communities he sees across the country. Phillips called PRN "the OG" and said other organizations copy its model. He said its monthly happy hours often draw hundreds of people, where most places would be lucky to get around 50, and that they are valuable for the industry's culture and for networking. He also emphasized that PRN now brings in industry-specific groups that create real business development opportunities. His example was a utility roundtable held the Monday before the recording. Hellbender and a couple of other companies in the audience took part, and he called it "a really prime field for us to get some new work." He said PRN understands it is a cultural engine and also that companies need business to survive.

Griffin said he has made several important contacts through PRN events over the years. A convention-center event a few months earlier produced both customer and partner leads. Seegrid has also found local suppliers for parts and service through the network, which means it now does business with and supports the local community.

Smocer gave numbers. At his first startup, an investor told him he rates a company's chances higher when he sees several people who have worked together before. At Mine Vision Systems, Smocer told his team they should not need recruiters in a community that is among the best robotics centers in the country, and that they should use their network and PRN instead. He said the company hired 18 people over the past year or so and that 14 of them had previously worked with someone already at the company. He gave first credit to his team's advocacy and called PRN "1B." His VP of engineering asked for a PRN membership within three months of joining.

41:35

Advice for Manufacturers Still on the Sidelines

In a rapid-fire round, Luecke asked how a manufacturer that has been watching from the sidelines should begin. Griffin's first answer was to "get started immediately," because autonomy is a journey that takes time. His second was to "start with a win." That means choosing an achievable, low-risk application that causes little operational disruption and needs little change management with employees, then adding a second win and increasing complexity and significance from there.

Phillips agreed and added that companies should understand the ROI calculation thoroughly. He said many missteps come from chasing the buzzword without understanding the impact. His advice was to start with a limited scope, get it right, and use the ROI analysis to set that scope. Luecke asked why Phillips said "it's never been easier." Phillips pointed to vendors like the three on stage. He also said autonomy was true R&D ten years ago and "sort of a pipe dream" twenty years ago. Today, he said, the question is whether you can identify the right use case and execute, and execution often "comes down to change management more than it does the technology."

Smocer cautioned about balance. "If you don't know what to do, do something" becomes reckless if done badly, and insisting on a complete strategy can lead to endless analysis with no action. A company can't be "a slave to either one." Applications differ, so wins have to be chosen strategically. He reminded buyers that adopting autonomy is a transformation and that employees understand their careers are affected, so leaders have to help people through it. His last point was about partners. They do not need to have everything figured out today, but they should be credible, have been through similar work before, have a vision, and be willing to engage with the customer's perspective. Because "nothing is completely figured out" and everything is moving fast, he advised choosing partners who will be around for multiple years.

45:39

Pittsburgh's Next Chapter: Success Stories, Exits, and Telling the Story

For the closing question on Pittsburgh's future role, Smocer came back to success stories. He said he joined the startup community to help create a virtuous cycle with local talent. He wants everyone, whatever their role, to understand they represent a community and to aim for outcomes that are visible and measurable in business terms, not only technical ones.

Griffin made the point more concrete. In his view, success stories lead to capital-injecting exits. Those exits fund the next round of angel investors and serial entrepreneurs, and the mid-to-upper-level leaders of successful companies become senior executives at the next wave of Pittsburgh startups. More capital, confidence, and capability build on each other, which he called "a self-fulfilling prophecy over time." He said Pittsburgh has not yet seen this at the scale he expects in the next few years.

Phillips said Pittsburgh is at "a very unique point in time." Drawing on the city's steel history, he described an opportunity to be "the vehicle for the physical AI world," with autonomy as one vertical within it. He saw the main challenge as cultural. The region has the engineering and manufacturing talent, he said, but is "a little too Midwestern humble." In his view, pride in being gritty often means being quiet and showing up to work with a lunch pail. He said Pittsburgh could learn from how Silicon Valley, Boston, and New York sell a story and a vision, and should talk about its strengths knowing it can back them up. He pointed out that Silicon Valley already knows about Pittsburgh: "They come here and they buy companies and they hire people." The remaining task, he said, is to convince the rest of the world, especially non-technical business people, that "Pittsburgh is where you go to build autonomy, to build the physical AI world."