ElectrifAi builds pre-structured ML models for AWS SageMaker

  • November 18, 2020
  • Steve Rogerson

New Jersey-based practical artificial intelligence (AI) and pre-built machine learning (ML) company ElectrifAi is releasing a collection of pre-trained and pre-structured ML models for Amazon SageMaker to sell.

SageMaker is a fully managed service from Amazon Web Services (AWS) that provides developers and data scientists with the ability to build, train and deploy ML models quickly.

The 36 pre-trained ML models have been added to AWS Marketplace. Clients can see results in days and weeks as opposed to taking 12 to 18 months since they turn data into actionable insights.

ElectrifAi has domain expertise across several verticals including banking, financial services, insurance, communications, media, entertainment, healthcare, consumer and retail.

AWS Marketplace users can unlock insights on data using that ML models that are based on industry specific use cases.

“We’ve been doing machine learning for 16-plus years and have deep expertise across many verticals,” said ElectrifAi chief executive officer Ed Scott. “Our machine-learning models help our clients drive rapid cost, profit and performance improvement.”

Through discovery conversations, the company can help identify business problems that its clients are trying to solve. After initial discussions and analysing the data, ElectrifAi determines the appropriate use-case pre-trained model that most accurately fits the requirements. Once a common data model is provided, ElectrifAi trains the model with the client’s data in their own environment to generate the signals related to the expected outcome.

No model is 100% accurate, but the model can be hyper-tuned to fit the goals and risk tolerance. ElectrifAi calls such models pre-structured, as they require little modification and tuning. These pre-structured models navigate the nuances that vary from client to client within a given industry.

For churn mitigation as an example, ElectrifAi applies several pre-structured models for segmentation and propensity modelling.

“We have a process we call the machine-learning model factory with pre-trained, pre-structured and brand-new models made to address client specific pain points,” said Luming Wang, CTO of ElectrifAi. “The model is explainable and tests for bias, resulting in easy-to-understand actionable insights that help solve real business problems.”

ML automates what people do manually when looking at large volumes of data, such as forecasting revenue, predicting demands, identifying and reducing customer churn, identifying high-quality customer prospects, fraud detection, risk management, analytics, and natural-language-processing. These ML models can help companies achieve a digital transformation. ElectrifAi has done the hard work already. All that is left to do is run the data through these pre-trained or pre-structured ML models in the client’s environment and integrate the output into the business applications.

“When building a model for your company specific to an industry, we may already have some of the parts that we can provide very quickly with a lot of transparency so you can build on top of that and get where you’re going much faster,” said Jim McGowan, head of product at ElectrifAi. “You would also have the support of an additional team of data scientists who can help with re-training or customisations of these models. We’re not just selling software, because we understand the domain and can build the models to solve for an end business use case. “

Founded in 2004, ElectrifAi has around 200 data scientists, software engineers and employees. The company is headquartered in Jersey City, with offices in Shanghai and New Delhi.