Sramana Mitra: You’ve created all these heuristics and you categorized on the basis of that. How many such heuristics do you have?
Ashutosh Garg: That’s the beauty of AI. You can’t do it manually. Today, there’s a million different roles in the industry. We can’t do it manually. We use deep learning to model these things automatically.
Sramana Mitra: Tell me how.
Ashutosh Garg: You can take any individual in this world. What you will find is that no two individuals are identical in terms of their career, but all of us have shared journeys.
For every individual, we look at other people who are like this individual a few years ahead of them and use that to infer what this person has a potential to do. This person can not only continue doing what they are doing today but also has the potential to do five others.
Let’s say we have people who are quantitative analysts in a financial institution in New York and are two years in their job. I’m trying to predict what this person can do next in their career. What I see is that they either switch roles and move into data science because they have a similar background, or they tend to stay and grow as a manager. This typically happens at the two-year mark.
I’ve identified two different roles. They can either be a senior quantitative analyst or they will become a data scientist. There’s an enterprise which has an open position for a data scientist. On the one hand, there are only so many people in the world who are data scientists.
Now I can look at the quantitative analysts who are two years into their jobs and see if they might be interested. I can further look at this person’s profile and say, “This person played basketball and a lot of team sports.” They are likely to be more social. If they were a chess player, they are very analytical, and they are likely to go into engineering. This is all happening automatically.
Sramana Mitra: Associations are being determined by the algorithm.
Ashutosh Garg: That is correct.
Sramana Mitra: How do you seed the algorithm so that the algorithm can start drawing these associations?
Ashutosh Garg: Let’s take my example. Ashutosh did his education in mathematics and science. From 1993 to 1997, he was in college. He did electrical engineering. Maybe he’s technically okay; not necessarily phenomenal. He likes to tinker around. After that, he moved to the US and studied machine learning.
Each of these become nodes in my graph and there’s a transition happening. There are so many people from IIT Delhi. Some people went to do management. Some people came to the US. Post-PhD, he could have gone into research. He could have gone into academia. He could have gone into industry.
Each of these becomes a node and there’s a transition happening over there. Each node has a number of features from when someone is doing what, how long they are staying in that node, and how they are growing in the same node. What I can do is I can take this journey until 2008.
In 2008, he left Google to start BloomReach. Can I predict that he will do BloomReach? What you can do is keep removing a slice from my life and try to predict that slice based on everything else about me. You already have all the data.
This segment is part 3 in the series : Thought Leaders in Artificial Intelligence: Ashutosh Garg, CEO of Eightfold.ai
1 2 3 4 5 6