This is Part Two of a series on the skilling structure at iMerit, where Chief People Development Officer Anindya Chattopadhyay shares his insights.
iMerit trainees undergo an induction program that introduces them to emerging areas such as Machine Learning, Computer Vision, Natural Language Processing, sentiment analysis, e-commerce, social media, and geospatial technologies. Formative understanding of these domains is strengthened through hands-on data annotation tasks conducted under the supervision of L&D facilitators. This familiarizes them with service delivery concepts such as task instructions, quality, and throughput. Post induction, trainees are assessed and successful employees are placed in client projects for on-job training. For example, those in finance projects are trained on financial terminology and documents. They can then perform services like financial data extraction, corporate classification, and sentiment analysis. Other domains include medical and retail, where the subject matter experts provide insights and context.
Q: What is the average time period required for this skilling to be completed?
The skilling process is pretty long and is broken into four to five different pieces.
(The initial stages of skilling done by Anudip and during the on-job training were detailed in Part 1 of this series.)
After they complete the on-job training, and have the big picture idea of big data becoming very curated and kind of structured data for outside users, they then have an idea of what they need to do. The next step is the doing. At the time of starting with the tasks, the person might be performing at 60 to 70%, and then we gradually help them to get to the next 30 to 40%, and improve their quality.
The fine-tuning part is done through basic crowd based tasks on some crowd platforms. However we actively oversee the quality of the work and give feedback which helps them improve. At the same time there is a sense of seriousness and feeling of accomplishment because it is a live project. This phase may last for a month.
Once an employee is mature, she is put on harder projects, and gets support from the training team on the floor while delivering the work. We use custom iMerit tools or tools given by the client. With iMerit tools, we use them even while practicing during the initial training. But if it is a client specific tool, it can only be used during production. So we have a unique technique of helping people on the floor, in a very agile kind of training. The moment the person submits the task or even before that, the person can be helped. So they learn while doing the task. In iMerit the more the person is exposed to that work environment and actually doing the task, the more skilled they become.
Skilling comes in to help the person identify and come up with some analysis and insight based on the information – where does this work fit in, who is the user, and if a mistake is made, what is the impact, what is the context.
But there are also two different things we need to remember. One is if you just do the tasks as you are supposed to do, and have no connection to what you do. Then it becomes simple data in your mind. Skilling comes in to help the person identify and come up with some analysis and insight based on the information – where does this work fit in, who is the user, and if a mistake is made, what is the impact, what is the context. Once a person is properly skilled, then doing the work becomes part of logical decision making. When they get connected to what they do and the outcome, they feel that they’re part of something very fascinating and exciting and have a futuristic job.
Q: What are some skills that are very unique and particular to data labeling work?
In iMerit, we have so many different domains and each domain has a very unique kind of skill requirement. So if I broadly classify the entire structure – we work with image, text or audio information.
For image, I need a person to very patiently observe everything, remember their instructions and have computer skills and understand the shortcut keys that can make them super productive. The data labeler should have a sense of understanding of the quality, the kind of concerns if a submission is wrong, and maturity to complete what is expected.
When we start working on NLP, this domain requires a fair amount of understanding of the contextual meaning of each and every word and a group of words. Now if I add to this finance as a separate domain, we require a subject knowledge of finance, financial terms, financial transactions, and financial information. Similarly, within healthcare, computer vision foundation skills are required but along with that, I require the person to understand biology as a whole. >
Basic things like punctuality, discipline, and sincerity are key because we work with very tight deadlines and with certain benchmarks of quality. To maintain levels of accuracy that can be as high as 99.9%, you have to be very sincere, very punctual and you have to be very attentive while submitting the task. If I have something which has to be delivered within a turnaround time of 15 minutes, then the team should have fantastic coordination and a sense of responsibility. So that kind of team feeling happens when they get into the project and start doing work in a live environment.
Basic things like punctuality, discipline, and sincerity are key because we work with very tight deadlines and with certain benchmarks of quality.
This is how the knowledge bank of iMerit is structured, across four levels.
Q: What is your process of preparing to work with a new cohort of trainees?
This is one area in which L&D is carrying out more process improvements. We have a very agile or on-demand kind of project skilling requirement that often happens in an ad-hoc manner.
We are actually trying to build up the Learning Management System (LMS) with all the information, every time we train people on the floor. We try to really preserve particular information about skilling or its components, so it can become a replicable way of training people. This is in the initial stage at the moment.
There are two different ways to train people. One is where I have a predefined set of skills that I want to train people in. The other depends on upon the very agile nature of work, which is continuously evolving and we need to adapt to new skills, new domains, and new regions. The evolving type of skilling is what I want to capture to make it scalable.
Q: What is an approximate split of these two types and how does it add up to the entire skilling picture?
Often after a person is on the floor, we do not really look for more classroom-oriented training. Gradually we will have a kind of self-paced, learning mode from the LMS where people can be motivated to do this in their leisure hours. This will not require a structured classroom-oriented training. It’s a virtual blended learning approach where they can have the help desk where the trainers or the experts are there to help them when they are learning as and when they feel like.
It’s a virtual blended learning approach where they can have the help desk where the trainers or the experts are there to help them when they are learning as and when they feel like.
So instead of asking them all to come to a session where training will happen and a trainer is dedicated for that, we want to make the training very agile and flexible, where the training has the motive of making people interested in learning by themselves. That is the way iMerit should go forward and exist because we work with a large workforce, which also works across multiple shifts.
If I ask them to join training sessions after eight hours of their shift, they do not really learn at that moment. They want to take some rest, and then get back to that mental state where new information can enter their brain. I always believe that I can train a person when the person has that training mode on. So I prefer them to go back home, have a meal, have some music or talk to some people, and take some hours of leisure. Then they can use their mobile phones to learn. We are making the LMS mobile-friendly so they can have mobile lessons. We gamify the material so that they can enjoy it and in that relaxed phase, they learn better. So this is the way we are building L&D at iMerit.
(The Q&A is continued in Part Three, where Anindya discusses the skilling process and the latest iterations being introduced within the L&D structure.)