Mainframe modernization to AI apps: USPTO reinvents itself

Applications of AI

The US Patent and Trade Office knows a lot about innovation. Registering and tracking new inventions has been the agency’s primary mission for over 220 years. But five years ago, he realized the company had a new challenge. It was to revamp the in-house IT system that had accumulated over the past 60 years.

The first moment of clarity for the USPTO came from Aug. 15 to Aug. 23, 2018, when the office’s proprietary Patent Application Search and Monitoring Database (PALM) went down. PALM supported his other USPTO systems for patent filing, searching, and payment of patent fees. During the suspension, patent applicants had to file their applications by other means, and the USPTO was forced to refund applicants’ paper filing fees. The outage made it clear that we needed to change our IT practices.

Two years later, CIO Jamie Holcombe was hired to further scale the organization’s digital transformation efforts. He hired his CTO, Stephan Mitchev, and other new IT leaders and set out to instill broader technical and organizational change.

Since then, the agency’s progress has been significant but slow, Holcombe said.

“It took about 18 to 22 months before we saw real results,” he said. “But when they see it, it builds momentum. Now I have to extend it to a bigger business, and that’s where I am now in my fourth year.”

Mainframe Modernization Eliminates ALGOL

The USPTO’s latest database update milestone was reached last month in its Trademark Distribution Information System (TDIS). The USPTO managed to separate that system from his 40-year-old mainframe application, the Trademark Register Application Monitoring (TRAM) system, written in ALGOL, a predecessor language of COBOL that dates back to the 1960s.

Three USPTO staff members still familiar with the ALGOL system helped the rest of the development team convert TRAM into a distributed computing app compatible with the Oracle database. Holcombe says TRAM includes more than 800 individual services, and the complexity doesn’t stop there.

“The old TRAM is actually synced with the new CRM system, so both had to be updated to ensure data quality and integrity,” he said. “That’s been our biggest focus this year and we can finally get that anchor off our necks.”

When that work is completed in September, Holcombe predicts database development teams will be able to move to a new way of working: agile and DevOps methodologies with infrastructure automation with Kubernetes and containers. For now, his IT team responsible for these database apps does some infrastructure and test automation via Puppet and Jenkins respectively, manually updating them on servers in the Iron Mountain colocation data center. To do.

DevSecOps and product thinking

The mainframe modernization team is an example of what happened with the USPTO’s human side of IT transformation. The institution has added a set of approximately 220 peer but segregated groups within its traditional hierarchical organizational chart. According to Holcomb, vestiges of the hierarchical system remain where necessary, but in which employees from different departments are assigned to smaller groups for short periods of time.

“It’s a team that can really do whatever it takes for that mission,” he said. ” [hierarchy] All of these different teams are responsible for securing resources. ”

These groups take full lifecycle ownership of each of the institution’s IT services and operate them like products throughout their lifecycle rather than as short-term projects. Spence Spencer, director of system configuration and delivery automation in his CIO’s office at the USPTO, said this approach is also known as product thinking, so the USPTO, among all other technological transitions, He said he was able to keep his IT resilient.

For example, Spencer’s team was able to respond in February to a critical security vulnerability that required the redeployment of application components. About 19 hours total lead to fix the vulnerability Part of his time was waiting for a window to update the production environment outside normal business hours, but the update was tested within hours It was deployed in the environment, Spence said.

“So the agency completed the entire software lifecycle, from conception to deployment, within 24 hours,” says Spence.

In the past, such changes could take up to six months, Holcombe said.

In addition to making fixes faster, Spencer said, product thinking among product teams encourages building resilience into the service from the start, including security features.

“That team knows they’re going to be living with this thing in five years,” he said. “They have a strong incentive to try to get it right.”

For new apps, the USPTO team adopts a DevSecOps approach, with features like static code analysis and vulnerability detection built into the GitLab CI/CD pipeline that automates deployment to AWS cloud infrastructure. I’m here. This prevents at least some vulnerable code from reaching production in the first place, Spencer said.

Tensorflow, ChatGPT Lead AI App Efforts

Mainframe modernization and DevSecOps are still underway, but the USPTO is also embarking on the next phase of digital transformation: automation with artificial intelligence. His group of 12 engineers at the USPTO had Google certified his TensorFlow open source machine learning platform two years before him, and the investment is already paying off, he said. said.

“When patents come in, they need to be classified, and we use contractors to [do it]said Holcomb. “But neural he’s also trained networks to classify, and he’s been able to free more than half of the humans that were doing the classification. That’s worth tens of millions of dollars.”

Next, the USPTO will investigate how ChatGPT can be used to augment examiners conducting prior art patent searches when evaluating new applications. According to Holcombe, tests so far have shown that the ChatGPT algorithm can personalize results for each examiner.

“Each inspector is always unique and the AI ​​is giving them what they want,” he said. “Remembering their choices and the relative nature of what they deem important.”

Beth Pariseau, Senior News Writer at TechTarget, is an award-winning veteran of IT journalism.she can be reached at [email protected] Or on Twitter @PariseauTT.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *