Q: Ratish, first question for you: What did you do before joining Roots Automation as Head of Artificial Intelligence and Machine Learning?
Ratish: Before joining Roots Automation, I was the Head of Data Science at a mental health startup in New York City called Quartet Health. They connect people to mental health care. I was an early hire there, so I had the opportunity to build the data science ecosystem from scratch and write algorithms to automate the end-to-end journey of a patient; algorithms like entity resolution, patient-provider matching, identifying unobserved mental health conditions, optical character recognition on forms etc. Many of our use-cases involved using natural language processing and computer vision to automate product and operations workflows.
In general, my background is in writing ML algorithms, building scalable machine learning systems, creating and growing data science teams in a fast-paced start-up environment.
Q: How did you fall into this space – into Data Science, AI, and Machine Learning? What was your journey like?
Ratish: I started my career as a programmer, but I always loved math, and wanted to work in a field where I could apply it. So, I naturally gravitated towards machine learning and went to grad school to study it. At the end of the program, I wanted to work for an AI/robotics startup. At school, one day, I was just bored and started Googling things like “Top 10 Startups in the Bay Area for Machine Learning.” I reached out to one of those few startups, had interviews and fortunately, ended up joining this small group (Wise.io) in Berkeley California, who were building AI/ML applications for enterprise customers. It was a great opportunity for me - I built machine learning models across various domains like lead scoring, churn, smart cars, content optimization, recommender systems, automated customer support etc. After working there for a couple of years, we were acquired by GE, which was quite pleasant for me.
I worked with GE for a couple of years, using computer vision to find cracks in oil and gas pipelines, and using natural language processing to build recommender systems for power plants. My journey in AI/ML has touched many different domains since, but Roots Automation reminds me a lot of my first AI job, and I am excited to be here.
Q: Are there any exciting projects that you’d like to talk about?
Ratish: Oh yes. One of the things that really excites me is building Digital Vision with the team here. It involves putting multiple large scale computer vision and language models together to build a cool product. The opportunities are endless on this project – we talk about active learning, generative modeling, simulation engines, reinforcement learning, etc.
Q: You mentioned that the space you’re in changes quite a bit. What are some of the big changes that you're keeping an eye on right now in this space, and how do you predict that the space will change within the next 5 to 10 years?
Ratish: The AI community is moving fast in solving the problem of natural language and computer vision. Big tech companies have hundreds of data scientists working around the clock to build new and improved machine learning models in that space. Every few months you hear Open.ai or Google or Microsoft hitting a new milestone, like beating the best chess player in the world without using any training data, or a language model that writes poetry, or a multimodal network that can play games, caption images and chat at the same time. Right now, the AI community is waiting for the release of GPT-4, which is expected to be the most advanced language model.
At Roots Automation, our job is not to solve the problem of vision or language, but to take these models, tune them, fit them to our problem space, and build new, cool products with them. The more that space advances, the better our products become, which is why we are always looking for the next big advances in NLP and computer vision.
Q: Last question: can you share anything that we're going to be working on in the next couple of months that will significantly impact our customers?
Ratish: You know, I’ve only been here a few months, but it seems like we have signed many customer deals recently and we are signing a lot more. This necessitates building a versatile machine learning system that scales across multiple customers very easily, supports different types of machine learning models (like computer vision and NLP applications), continuously learns and improves from data over time without much manual effort from data scientists. I am excited to build it alongside the brilliant AI/ML team here. I have built similar systems in the past, but the ML world changes so quickly and every year, there are new open-source libraries leveraging the latest and greatest in the space, so it will be a fun learning experience. In terms of impact on customers, this system will reduce time to go into production, improve reliability and performance of Digital Coworkers.
For more thought leadership on data science, artificial intelligence and machine learning, be sure to follow Roots Automation and Ratish Dalvi on LinkedIn, and subscribe to our blog!