AI-driven automation was supposed to replace ineffective and outdated business practices that really can’t support today’s real-time, global economy. While the actual implementation of automated applications has taken longer than initially expected, the leaders at Oii.ai say they've found a business area ripe for disruption through innovation.
For too long, too many companies that oversee extensive and complicated supply chains have relied on spreadsheets and other outdated tools, said Bob Rogers, who co-founded Oii.ai in 2019. The goal of Oii.ai is to help clients deploy AI systems that improve the efficiency, resiliency and ability to optimize — in near real time — those supply chains.
"The tooling that is used in the supply chain is really quite limited," Bob said. "It's not 21st century tooling, by and large. So there's an opportunity to take AI automation and use it to drive way better performance, both financially and in terms of service."
Like a lot of companies seeking to leverage AI, Oii.ai requires high-performance processors but Bob also said that all hardware employed must also be trusted to safeguard client data. After all, much of the data will come from disparate systems and even disparate companies and can be quite sensitive–so secure collaboration is essential. Like most companies, Oii.ai doesn’t wan tot have to invest in making the company's software developers become hardware experts. As such, Bob credited AMD processors for providing strong security, via AMD's Secure Encrypted Virtualization (SEV), which encrypts all CPU register contents when a virtual machine stops running. He also gave a shout to AMD for providing a highly effective "bridge" between software and hardware.
"Our developers and data scientists are working in PyTorch or JAX, and they're focusing on the code they're writing to solve the problem," Bob said. "That bridge to the hardware that allows the software to be optimized to work as well as it possibly can on that hardware is actually critical. That's a really powerful place for the hardware to play a role, because I know that the optimizations that AMD does to make these libraries work really well with AMD hardware."
Through Optii, the company's flagship software product, Oii.ai monitors data from supply chain systems. The data includes customer demand orders, how material flows through the chain, what's occurring with suppliers, lead times, and shipping. Users can identify bottlenecks and disruptions and learn what's causing them.
To illustrate the company's computational needs, Bob defined the three main areas of Oii.ai's service. The company simulates customers' supply chains, which means creating a lot of multi-threaded Monte Carlo simulations, the mathematical technique used to predict possible outcomes from uncertain events. This requires high-performance computing and efficient memory use, Bob said.
Next, Oii.ai employs AI machine learning and modeling components to understand the characteristics of a customer's supply network. It's not just modeling, but often also time-series modeling and is a very different computational workload. The third area is the Generative AI employed by Oii.ai. The GenAI enables customers, even non-technical personnel, to ask questions about their supply chain and receive answers with natural language.
Asked to provide advice for those ready to make the leap into GenAI, Bob said a lot of lessons were learned during the era of digital transformation, starting with diving into large transformation projects without a clear understanding of the potential ROI.
The "antidote" is to tackle a couple of small projects, this way, critical knowledge is obtained quickly and cost efficiently.
"One of the biggest things for technical efforts around AI is to have this architecture where the AI helps you marshal tools you already have, rather than asking the AI to give answers," Bob said. "One of the biggest challenges is hallucination… I'm excited about the advent of AI that hallucinates less. Right now, AI doesn't know anything. It just talks a good game. That's literally how AI works. But the minute AI actually knows things, that's going to give us a step change in our ability to compute things reliably."
