A recent Forrester report called data science and AI engineering company Tredence a ‘dark horse’. Founded by Shub Bhowmick, Shashank Dubey, and Sumit Mehra in 2012, Tredence focuses on the last-mile delivery of data science. The company caters to clients based in the US, Canada, Europe, and Southeast Asia. Tredence has launched a series of innovative solutions driven by the goal to ‘humanise, solutionize, and operationalise AI’. Case in point — MLOps platform, ML Works.
Analytics India Magazine caught up with Tredence co-founder Shashank Dubey to understand more about Tredence 2.0 journey and their unconventional approach to talent acquisition and upskilling.
Sign up for your weekly dose of what's up in emerging technology.
AIM: How has the analytics industry evolved over the years? What are the important trends you see?
Shashank Dubey: There are currently four major trends defining the industry: Clients want providers to shift from a project mindset to a product mindset. Data science is not just a project where you could build the most accurate algorithm and create a stack deck and dashboard around it. Clients want the ability to create full-stack applications around algorithms which can then be integrated into the decision-making process of the enterprise. For that, we need a very mature product mindset. Bigger clients want to build applications for their data science use cases and they expect their service providers to think like a custom product shop rather than a custom project shop.
The second big change was triggered by COVID. With respect to the revenue model, clients are expecting providers to have more skin in the game and not just work in a fixed cost world.
The agile way of working has been well-adopted in the IT side of the world by companies such as Amazon, Facebook, and Google. A lot of agile methodologies and philosophies are coming in data science as well. This means that there will be increased experimentation which will further solidify the product mindset. At Tredence, we have an Agile Centre of excellence where we have experts in the field of agile methodology, fairly unheard of in the data science industry.
Clients are also talking about operationalising data science. When we started, we talked about the last mile which involves three Ops — DataOps, MLOps, and AIOps. People could not understand what we were talking about. Now it has become mainstream. Last-mile also means having a strong operationalising mindset.
AIM: How did pandemic change the analytics industry?
Shashank Dubey: At present, people think more from the model design perspective when speaking about data science and analytics than about data engineering. Going forward, model design will not matter much because most algorithms will be available as APIs. In fact, companies like Tredence are building algorithms that have a high degree of verticalization across industries and can be made available as APIs. AI as API is a good differentiation. It allows data scientists to spend less time building algorithms from scratch.
Having said that, readily available algorithms can offer only up to 90% accuracy. The true test of a data scientist would be whether he/she can take the accuracy from 90 to 99%. It requires domain expertise, analytical thinking, and the ability to identify edge use cases.
Working around biases and long-tail use cases of AI systems would also become very important. While designing algorithms, data scientists often assume that the end-user is AI and not human. There is a need for humanising these systems. Design thinking has seeped into how software is built, next up it should enter AI algorithms.
AIM: Tell us about the transformative journey of Tredence 2.0.
Shashank Dubey: We have created a bunch of verticals and horizontals which are common in the IT industry. The difference is that, at Tredence, these verticals and horizontals are led by ‘mini CEOs’ who have complete autonomy over its functioning.
We are deepening our focus on developing a specialisation. We are hiring deep domain experts, consultants, and strategists. Good talent is expensive but it is a good investment. We would also be focussing more on developing and nurturing deep generalists in our current talent pool.
Channels and partnerships will define the future of data science, so one area that we are concentrating on is building robust channels that act as growth accelerators at Tredence. We have also invested heavily in building industry-leading accelerators. We have allocated a significant part of the investment towards the innovation council that provides clients a rapid prototyping platform.
AIM: What are the challenges Tredence faced while scaling?
Shashank Dubey: The biggest challenge is sourcing quality talent, which can work with the shifting landscape of technology. We need people who can reskill themselves.
It is more of a mindset challenge rather than a skillset problem. We are trying to solve this by propagating a deep generalist concept: the ability to understand and learn a piece of technology deep enough to create value. We have also realised that deep generalists are not naturally available in the market; the onus is on us to create a deep generalist career path for the talent we hire.
AIM: What are Tredence’s latest offerings in artificial intelligence and data science-based services?
Shashank Dubey: We are developing capabilities to move towards AI as an API paradigm. Our aim is to develop algorithms that go beyond vanilla data mining and understand the intent and semantics behind customer data. These multidimensional algorithms can understand the hidden intent and causal inferences behind common communication channels such as emails, chats, voice recordings, text messages. We are creating knowledge graph artefacts to create a semantic layer on enterprise data that facilitates higher interpretation of data.
Further, work is in progress in the Video analytics capabilities, especially in distraction detection. We are building algorithms for video analytics that go beyond CNN. It has implications for our factory shop floor where we use video signals to find human compliance.
Further, we recently launched our MLOps platform, ML Works. It is built using sophisticated algorithms to track other algorithms. ML Works offers automated workflows, pre-built solutions for tracking model degradation, and code workflow management. It will allow data scientists to shift focus from management and risk mitigation aspects of machine learning to driving innovations in AI.
Ops has three components — DataOps (making right and contextual data readily available), MLOps (productionalizing ML model), and ALOps (integrating AI algorithms to downstream processes for decision making). We are trying to create an end-to-end Ops platform, which will go a long way in realising our vision.
AIM: How is the company planning to use fresh investment?
Shashank Dubey: We have recently secured $30 million from private equity firm Chicago Pacific Partners. There are a couple of areas we would be focusing on:
- Hiring deep talents across big consulting/tech firms.
- Standing up a new vertical in healthcare. Value-based care has become mainstream
- Opening new talent centres in Chennai, NCR, and Toronto
- A part of the funds will be used for innovation, especially towards our Rapid prototyping factory and AI CoE.
- 12-13% funding has gone in returning the wealth to employees. We did an employee option stock buyback of about $3.5 million.
AIM: What are the major technology leadership initiatives at Tredence?
Shashank Dubey: We are getting into deep partnerships with companies like Snowflake and Databricks. Alongside, we have also built a full suite of data engineering accelerators, which perform functions from data ingestions to automatically read the data to build pipelines.
The ML Works platform is one of a kind in the industry; none of our peers built something as sophisticated. We are helping clients navigate through the multi-cloud paradigm. It is a task to manage assets across multiple clouds. Soon, we will be launching a CloudOps platform that will help clients optimise data assets across multiple cloud environments.
AIM: Tell us about the hiring process at Tredence.
Shashank Dubey: For lateral hiring, we take in people against specific skill requirements for the particular role.
Some of these requirements are:
Data scientist: Ability to build deep algorithms in Python, neural networks, and TensorFlow
Data engineering: Cloud-native ability, PySpark, Scala, building scalable data engineering pipeline, and expertise across clouds
When hiring fresh college graduates, we don’t necessarily look at specific tech skill sets but the idea is to make them a full stack analyst inside Tredence.
AIM: What are the upskilling programmes at Tredence?
Shashank Dubey: We have launched a Deep Generalist program for people with less than five years of experience. We give them very well-orchestrated training in their chosen path, which is spread over 2 to 5 years. The idea is to give them guidance in their choice of industries, horizontals, and skill segments. We align performance evaluations and appraisals accordingly.
Another initiative is the Everest program, which can also be called a mini MBA program, similar to a Bootcamp where people are trained on leadership skills.
We have also started partnering with various academic institutions, B-schools, engineering colleges, and other statistical institutes to help us improve our programmes and training offerings.