Nitrio’s NLP platform analyzes tons of email and phone conversations of sales representatives and provides sales managers with insights about sales performance, as well as offers advice to the sales reps, helping them improve their performance.

The core of the NLP platform was built using manual rules and heuristics-based models that provided satisfactory performance at the start but caused bottleneck issues as the company started to scale.

Squadex has developed a modern and robust intent extraction pipeline with TensorFlow. Continuous data annotation, training, and evaluation of deep neural network tasks have been automated, which allowed Nitrio to streamline onboarding of new tenants and to scale existing customer accounts.

The new Machine Learning platform has increased the daily throughput by 50% and reduced the amount of manual work 5x to develop and debug regex. This allowed Nitrio to optimize the cost of operations by 20%.

Increased by 50%

Daily Throughput

Reduced 5x

Manual Work

Optimized by 20%

Cost of Operations

Squadex Machine Learning Services

Research & Solution Design

Natural Language Processing Solution

Automated Machine Learning Platform

About Nitrio

Nitrio brings state-of-the-art data science directly to business users with a data-driven solution for the sales process optimization.

Nitrio’s technology harnesses proprietary machine learning and data mining frameworks and combines their customers’ data with billions of external data points to surface insights that revolutionize the way their customers sell.

PROBLEM

Nitrio analyzes inbound messages and rep-to-lead correspondence while collecting useful statistics for every sales representative. Nitrio provides their clients with data-proven buy-in strategies and keeps their top-level managers updated with the statistics about every sales rep’s performance, including common mistakes and errors they make. The inbound client emails can have various types of sentiment that have to be classified and tackled case by case. To handle the data of such specificity, the accuracy of sentiment analysis of no less than 95% is required.

The core of Nitrio’s NLP platform was built using manual rules and heuristics-based models. All messages that could not be classified by the platform were outsourced to a third-party company for manual processing, which increased service costs. Performance at the start was acceptable but became a bottleneck as the company started to scale.

SOLUTION

Squadex developed a machine learning platform that augments manually developed regex, aiming to increase the overall operational efficiency. The ML-based platform analyzes and classifies inbound messages—those that cannot be processed correctly are sent to human employees for review. By doing so, the platform greatly increases the overall amount of messages processed per day and reduces the number of manual tasks.

The machine learning platform is a neural network for natural language processing, which is developed using TensorFlow. Amazon SQS is used to get or send messages from or to the new ML-based platform. The ML platform is located on a separate EC2 instance, and it works based on hydrosphere.io. Continuous monitoring was built with CloudWatch.

AWS Products Used

Results

Nitrio received an ML-based platform, which efficiently analyzes inbound messages and classifies their contents. The new ML platform increased the accuracy of the existing system’s data processing capability by 50%, which allowed to reduce the amount of manual work 5x and accelerate the analysis of inbound messages. This allowed sales reps to shift their focus from processing emails to managing leads who are most interested in a company’s products and services. As a result of the implementation, Nitrio was able to enhance its product performance, optimize the cost of operations by 20%, and generate more revenue for the business.

Testimonials

The legacy platform that we had just wasn’t getting it done. Not only was our ML pipeline way too long, but it was costing us a lot of money since we’re wasting so much time on re-running the work we did in Spark. Squadex came in and optimized everything. They saved us a lot of money, provided us with an instant ROI and drastically reduced our release time. They really hit it out of the park.

Lenny Blyukher,VP of Engineering at NinthDecimal

Looking to Improve Your Machine Learning Capabilities?Schedule Free Consulting Session