Infrrd’s Machine Learning-Based Algorithms Extract Complex, Unstructured Data To Derive Deep Insights
Eliezer Yudkowsky, an American AI researcher and writer, says, “By far, the greatest danger of Artificial Intelligence is that people conclude too early that they understand it.”
When artificial intelligence messes up, it can be unsalvageable, or downright hilarious, as can be seen from the mismatched captions on these photographs tagged by one of the most advanced neural networks in the world — Microsoft Azure. Image recognition algorithms can make really bizarre mistakes — mistaking orange sheep for flowers, sheep on trees for birds, and sticks for giraffes. And we’re just talking about image recognition here.
For AI to be on point, it needs to be deep and accurate. Not an easy game as we can see. Amit Jnagal, the founder of San Francisco-based enterprise AI startup Infrrd, seems to be aware of the danger Eliezer warns about.
“True AI systems pass the test if you can’t make out whether you are dealing with a human or a bot. I think none of the systems are there yet. However, we have learned to intersect the technologies in the AI universe layer over layer to get more customised and deep learning solutions for enterprises,” Amit told Inc42 in a recent interaction.
He is also pretty confident his company is on the right track with the AI solutions its developing. Infrrd is developing advanced algorithms used in intelligent data capture to improve the data processing, especially for financial services companies.
Amit has quite astutely set his sights on a market that’s fast-expanding— enterprise solutions. Since AI research first gained impetus in 1956, it has moved beyond robotic applications to automating the daily processes in enterprises at large scale. Today, enterprise AI is a globally recognised industry finding applications from retail to ecommerce to finance and beyond, with its market value projected to increase from $202.5 Mn in 2015 to $11.1 Bn by 2024.
A Look At Infrrd’s AI Platform, Target Market, And More
Launched in October 2016, Infrrd terms itself a ‘machine intelligence’ startup that utilises computer vision, natural language processing (NLP), and a predictive algorithm-focused artificial intelligence (AI) platform to develop enterprise solutions. The machine learning-based algorithms it has developed extract complex data from images, bulk documents, etc and derive deep insights from big data to help enterprises automate tasks and solve different business problems.
“We see huge potential in automating human tasks that need low-level cognitive capabilities like reading a document, looking at forms, understanding clauses of contracts, extracting data from images, and more. We have helped a lot of our customers cut down their manual data processing spend by over 70%,” says Amit.
Amit explains that a lot of ongoing AI initiatives in the financial services space that are focused on fraud prevention, anomaly detection, etc, are primarily working on structured data coming from customer interactions.
However, Infrrd does things differently. “We use the same algorithms that some of the world’s leading companies are using to build text extraction and analytics solutions, but direct them to solve a more immediate need of automated data processing,” says Amit.
Infrrd mainly caters to the medium-to-large scale financial services industry. The startup is working currently with more than 35 clients globally, including Fortune 500 companies in segments such as retail, finance and real estate verticals.
After bootstrapping for more than a-year-and-half now, and making the US and UK Infrrd’s primary market, the company is now steadily trying to deepen the presence in India and is further looking to enter into the European markets.
Infrrd: Offering AI-As-A-Service And Customised Big Data Insights
As Amit says, there was a time when documents and images were transformed into a structured text format manually. Not only was this time-consuming and cumbersome, but the process was also low on accuracy. Over the years, technology has evolved in this domain and we now have products like IBM Watson, Google Vision, Amazon Rekognition etc.
However, with most algorithms being aligned to work across all sectors, the results are generic in nature. For instance, when a retail customer tries to use Google Vision for identifying the features of a dress, it tells her/him that it looks like a dress and provides some basic details. Most enterprises already know this and they want to know something deeper than “this looks like a dress”. Things like: Does this dress have a distressed look? Can it be worn in summer? Is it a strapless dress? Does it have ruffles?
“This is what we try to extract from images,” says Amit. “We have three fundamental capabilities — NLP, computer vision and predictive analytics — which help us offer our clients an AI-as-a-Service model with required customisation on a long-term basis,” he adds.
Here are some of the products and solutions offered on the Infrrd platform.
The most popular of these offerings is Infrrd OCR (optical character recognition) — intelligent data extraction for complex documents. “A few years ago, it was not possible to automatically process documents that had fairly complex data elements like high-density tables or complex legal clauses. We are one of the first companies to use cognitive capabilities like computer vision, deep learning, and NLP to read and build context around documents,” Amit explains.
Creating A Differentiated Edge In Enterprise Solutions
While Infrrd may have outpaced the traditional OCR and data extraction solutions, it now has a rising brigade of enterprise AI startups in different corners of the world to compete with. In India, there are startups like Fractal Analytics, Artivatic, Mad Street Den, SigTuple, and Staqu among others in the artificial Intelligence space.
Being a global startup, Infrrd is competing with extremely large horizontal AI & RPA systems like Kryon, Workfusion, Automation Anywhere, etc and vertical offering specialists like image recognition companies or NLP platforms like IBM Datacap, Nuance and Cvision.
However, Amit says that since the company focuses on just a couple of verticals rather than providing a wide range of solutions, it is able to provide extremely deep and high-accuracy solutions. “Our fast time to market also helps us differentiate between the players in the first category,” he adds.
He also claims that Infrrd is one of the few companies that can combine NLP and image recognition capabilities in a single solution, which helps it differentiate from vertical technology players.
Amit also says the pricing offered by Infrrd is very competitive. “We can extract data from documents that most other providers cannot, and if they can, it comes at a high price,” he says. He explains that Infrrd’s processes and platforms have been built from the ground up to accommodate for 10-15% customisation for its customers. “A one-time cost is charged for customisation and then we offer a SaaS-based solution, which companies keep for two to five years on an average,” he adds.
This customisation helps Infrrd go the extra mile for each customer and deliver solutions that have the feel of a tailored solution. Additionally, Infrrd also learns from its customer interactions as they use its platform and automatically corrects any accuracy issues.
“We initially source data from a company’s past records or the publically available data from our global repositories. With time, however, the platform learns more from the ongoing interactions. This helps us deliver a better experience for our customers and ensures their satisfaction,” he adds.
Infrrd Aims To Be The World’s Best Intelligent Data Extraction Company
The world has just started zeroing in on AI on a mass scale and the industry is far from saturation. Research company Tractica forecasts that the revenue generated from the direct and indirect application of AI-based software is estimated to grow from $643.7 Mn in 2016 to $36.8 Bn by 2025. This represents a significant growth curve for the nine-year period with a compound annual growth rate (CAGR) of 56.8%, as the report mentions.
But as Vivek Wadhwa, a Distinguished Fellow and professor at Carnegie Mellon University Engineering and Harvard University and a director of research at Center for Entrepreneurship and Research Commercialisation at Duke, cited Duke professor Dan Ariely recently, “Artificial intelligence is like teenage sex. Everyone talks about it, nobody really knows how to do it, everyone thinks everyone else is doing it, so everyone claims they are doing it.”
Apart from finding their niche in the broader spectrum of AI, this is another challenge startups are going through — to prove their worthiness and claim a product market fit.
So far, Infrrd has been able to do both and has successfully run the show without any external funding. But in global markets. As of now, 95% of its revenue comes from the US and the UK. The Indian market, owing to its vernacular language content, poses a huge challenge for such startups; besides, it’s price sensitive. But Amit seems optimistic. “India might be price sensitive, but it offers volume, which makes up for the overall price. We are working on our vernacular content offerings as well.”
“Our purpose is to build world’s best intelligent data extraction company and we are well on the way to get there,” he signs off.