At Wisedocs, we’re building a stellar team as we continue on our accelerated growth path. In this new employee spotlight series, we will be highlighting our Wisedocs employees and getting to know them beyond their job title - and also what it’s like to be in their role. In this Brain Power feature we meet Rajiv Abraham, our Lead Machine Learning Engineer at Wisedocs. Rajiv joined Wisedocs in June 2020, when they were originally Bear Health Technologies, after spending time across startups and major global organizations.

Why is machine learning important to Wisedocs?

Wisedocs uses machine learning models to build their intelligent document processing platform. Our machine learning processing of documentation allows for high rates of accuracy, even on previously-unknown documents.​ Manual document processing can result in human error and inefficiencies - claims professionals spend nearly 50% of their time on activities that don’t impact the outcome of the claim. Wisedocs is here to make efficient document processing a reality and machine learning and artificial intelligence enables us to get there.

Beyond that, indexing, organization, keyword pull, and deduplication are other features made possible by training a machine learning model on existing records to apply its intelligence to records it has not seen before and provide the most value, time saved, and cost savings to our clients which ultimately benefits the claimants and the industry as a whole.

What is a machine learning engineer?

A machine learning engineer (ML engineer) is a technical person who focuses on researching, building, and designing self-running artificial intelligence (AI) systems to automate predictive models. Machine learning engineers design and create the AI algorithms capable of learning and making predictions that define machine learning (ML). Now, let’s jump into the interview and meet Wisedocs’ very own machine learning engineer, Rajiv!

Wisedocs introduces: Rajiv Abraham

“Hey everyone! I’m Rajiv. I have a passion for software engineering with specific interests in integrated development environments (IDEs), compilers, domain specific languages or programming languages and databases. I like thinking about building a language workbench and a database for machine learning (ML). I used to run a meetup group for functional programming in Toronto (similar programming groups can be found on Meetup for developers) for almost 3 years which was a lot of fun.”

Tell us about your career. What brought you to this point?

“My first job was at a small company in Mumbai, India which served clients in the US. I was a junior developer. Due to high employee attrition, I got to lead a team of 4 senior developers. It was hard at first, but became fun once they accepted me.

“I realized that I had not given my best in my undergrad program back in Mumbai and resolved to work hard if given a second chance to study. So, I did my Master’s program in Computer Science at Concordia University, Montreal.

“From there, I worked at SAP Labs in Montreal for 5 years, theScore at Toronto among others before joining Wisedocs.”

What made you interested in technology and machine learning?  

“I didn’t want to work in any of the other engineering streams 😊 (such as,mechanical or chemical). I did my undergraduate degree in Computer Engineering and kept on going. I didn’t necessarily seek startups, but I wanted interesting work, whether startups or otherwise.

“In my most recent job search, I told Leo Zovic, the lead developer at Wisedocs at that time that I was looking and he suggested Wisedocs.”

What gets you excited about working at Wisedocs?

“I like thinking of building reusable software libraries and systems for ML. I’m totally amazed at how the complexity of document indexing stretches my design assumptions or even breaks them.

“We just don’t have one or two models here. We have to think about numerous models interacting with each other in complex ways to give us the final outputs. Building systems to make ML at scale manageable is fun and creative!

“I’ve always been interested in the intersection of ML and Healthcare and Wisedocs is the perfect place for that. We have a treasure trove of documents for ML. The applications that we can build over those which interplay with each other can be quite exciting to think about.”

At Wisedocs, we know that our people want to feel challenged and connected to the work they do. This is what makes solving these complex problems so engaging for our employees across all teams.

Share about the day-to-day of Wisedocs' lead machine learning engineer

“For me, an ideal day would start with assimilating all my “To Dos” and ideas that strike me the previous day and compile a task list for the day.

“My current strategy is to unblock others first. I used to have focus times first thing in the morning and end of day and I’d love to come back to it. We have a stand up meeting at 10:30 AM.”

Rajiv’s day-to-day consists of:

• Coordinating and ensuring that team members are unblocked in their individual functions and tasks plus making sure that they have enough sense of the entire system to work on their focused task and an understanding of how it fits within the entire ecosystem of the company.
• Anticipating, communicating, and mitigating the impact of new features and their ripple effects on our systems.
• Anticipating what challenges my team may face or even seeking feedback from them on how to improve the system and development experience.
• Breaking down the product team’s requests into concrete tasks and tickets for my team to work on.

The impact of a manager makes up 70% of the variance in a team’s engagement, according to Gallup. At Wisedocs, our managers know the importance of supporting their teams. Whether they are coaching, leading, or providing subject matter expertise to their teams, our managers are always wearing many hats.

‍

What’s your stack as a lead machine learning engineer?

“I’m quite excited about the Ray ecosystem, which is a distributed system in Python land geared towards ML. I also like the training platform Determined.AI, which is a fast way to train and build deep learning models."

Our stack at Wisedocs is:

• ‍‍Tensorflow
• ‍PyTorch‍
• Python

“As we scale, I’m so looking forward to bringing in more technologies and systems to speed us up.”

What are the perks and challenges of being a lead?

“A perk is having a say in the design and execution of building the ML Platform here at Wisedocs. This can be creative and satisfying and I’m looking forward to implementing them.  On the other hand, a challenge is blocking focus time to still do my concrete development tasks in the midst of all the meetings. As I mentioned before, I try to start the day earlier than others so I can get a block of time where I can do focused work.”

Time blocking, also known as chunking, is one of the most important productivity methods you can implement to protect your work and calendar. It involves designating certain times of the day to work on specific projects or tasks. Working in timeblocks helps you prioritize your work and minimize the cognitive load that results from task switching and multitasking.

What part of the Wisedocs product are you most proud to have worked on, to date?

“I’m proud of introducing a data oriented approach to evaluating the ML models. We now have granular metrics on how accurate our models are even for the attributes for each document. I’m proud of making a generic framework for adding heuristics for structured documents. I was quite happy with the heuristics/model framework that I built for one of the projects.”

Feeling like the quality of your work is strong and recognized by colleagues helps build a sense of pride at work. Having pride in what we do at work is another important component that contributes to our engaging workplace at Wisedocs. We’re creating an environment where we all have a measurable impact on the problems we are solving.

What are the best steps to becoming a lead machine learning engineer like you?

“I think a tech lead has to shift focus from building components to building systems (.i.e. the relationship between components). This means upgrading one’s knowledge on system design and architecture. This Github page on designing large scale systems is a good start.

“One also has to find time to learn, research and prototype new technologies again at the systems level (as opposed to learning small cool libraries used within a component). So when your product team comes to you with a scaling problem or complex task, hopefully, you have a suite of tools and frameworks that you can think of to solve that problem.”

Take your next career step with Wisedocs

Rajiv is just one of the bright individuals that has chosen to build their career at Wisedocs and his machine learning team plus many other teams are always looking to hire new roles. If you want to make a big impact on shaping the industry of intelligent document processing, then a career with Wisedocs may be for you! We encourage you to check out our open roles and apply via the Wisedocs careers page. We are always looking for great people to join our team!