“If only one of us were a teen counselor, we would be able to effectively train the machine learning algorithm and learn the patterns of cyber bullying within an ecosystem,” wishes Birago Jones. “Precisely, and then, use the same algorithm to build data sets, which could give us insights by predicting possible outcomes within that confined environment,” adds Karthik Dinakar. There was a momentary pause where the two MIT scientists looked at each other, spellbound by the revelation that occurred to them at that moment. In that split-second, Jones and Dinakar found the missing link between contrasting domains such as teen counseling and machine learning: Enabling domain experts to train the machine learning algorithms, instead of outsourcing it to a programmer. Working on this proof of concept, years later, the duo went on to establish a company—Pienso—around the proposition and extended the same idea to various other industries.
Today, Pienso stands as a ‘sui generis’ entity that enables non-programmer and industrialists without any machine learning proficiencies to tailor machine learning algorithms in accordance with specific data sets and enhance data-driven applications for their domains. Jones explains how companies seeking machine learning capabilities do not necessarily have hands-on involvement in the development of machine learning-driven applications. Quite often, either the perspectives of domain experts and programmers lack parity or the development of applications takes a substantial amount of time due to reasons such as iterative cross checking processes.
Pienso serves as a catalyst that facilitates the amalgamation of domain expertise and machine learning
“Pienso employs a technique called, ‘lensing,’ which figuratively brings out the domain expert’s perspective in the machine learning-driven application, allowing the expert to aid the development of the machine learning algorithm from the initial stages,” explains Dinakar.
The company consolidates its offerings through an API driven self-service platform that allows organizations to capture data, interact with machine learning algorithms, carry out multiple iterations of development, and churn out the essence of the envisioned perspective. As a case in point, consider a cardiologist seeking an intelligent dataset. Pienso emphasizes that a cardiologist is the apt person to train the machine learning algorithms to achieve the desired outcome, owing to her immense experience in the field. The cardiologist can tinker with the algorithm at her expense instead of waiting for a programmer to make any necessary alterations. This methodology not only eliminates the iterative stages of upgradation but also saves time and resources. The entire progression serves as a classical example of sourcing craftsmanship from the in-house expertise.
As technology enablers, Pienso serves as a catalyst that facilitates the amalgamation of domain expertise and machine learning; a recent endeavor of the company sheds light on this process. Pienso was bestowed with a herculean task of organizing five year’s worth of data, distributed over millions of documents. By empowering the knowledge management team at their client’s premise, Pienso aided the creation of a machine learning model that analyzed every document fed to it, classifying it according to various categories such as authorship, genre, priority, and ranking. By implementing filtering criteria, the machine learning model was able to segregate documents based on search results and categories to which the documents belong. What would generally require months of development was accomplished within days using Pienso’s machine learning platform.
The company has managed various such projects that serve as worthy testimonies for their motto “democratizing machine learning.” By empowering domain experts from various industries to take initiatives in the machine learning, Pienso is continuously adding value by nurturing ownership within their clients.