Infrrd Labs Tackles Tough Problems

Infrrd Research Labs keeps innovation going. Just ask Pratyusha Rasamsetty, who runs the R&D department, a subset of the 200 person engineering, machine learning, and AI group. 

“It’s very important to have a department like this, to focus only on innovation and to make sure that the company comes out with new ideas,” she says, adding that having the ability to explore problems without a specific timeline or customer pressure will give the company an advantage in the long run.

Rasamsetty got involved with the company in an earlier iteration a decade ago and started Infrrd Labs in July. Teams within Labs meet weekly, combining knowledge across related topics to explore problems that need solving and that could benefit Infrrd’s product line.

Creating More Customer Value, Faster

Divided into two groups, the teams brainstorm around the latest research in AI and machine learning, working off of newly published papers to develop better solutions. 

“We can focus on the best way to solve the problem,” she says. “We are trying to find out the best possible algorithm so it’ll give the best results at the end.” 

All of this learning could lead the company to a place where it doesn’t need training data to get customers started, she says. So while it could still be fine-tuning solutions to improve extraction accuracy, customers or potential Infrrd clients would be able to try the product quickly, without providing any data. 

Solving The Toughest Challenges

Alongside industry giants, Infrrd Labs is working on tackling tough problems. Its focus is currently on key-value pairs, to better map data representations -- such as mapping ‘Dallas’ to ‘City’, irrespective of the document type. The company looks to build solutions that will be useful across customers and domains. It’s also trying to build models to account for common fields from the get-go, to further reduce the reliance on training data. 

“As part of these Labs, we try to build generic problems which can be solved for anybody, not just one particular customer or domain. We try to glean useful information or approaches from that,” she says. “We have seen some good progress happening.”

Infrrd Research Labs plans to expand its team and continue adding new problems to solve. 

Working On Big Innovations For Our Customers

Meanwhile, Infrrd’s current customers and potential customers alike come out ahead as the model becomes more mature, and the need for training data, which needs to be stored, diminishes. Reducing the high dependence on data means fewer data storage and governance considerations, as well as a more nimble model, she explains. “Right now we promise customers that we have a feedback loop built into the system which takes corrections as the machine learns on its own, and the models get better every time someone feeds something into the system.”

Going forward, she says she’d like to see the current team’s skills and approaches get stronger, even going so far as to submit their own papers and patents with their own technologies. Rasamsetty adds that she’s thrilled to be part of the forward motion as Infrrd continues to innovate for customers on their digital transformation journey.

“Establishing this department means we’ll know before something is coming out,” she says. “Or we might be able to come out with the next big innovation in the AI world.”