At Brillio, the opportunities that I have gotten have enabled me to refine the ART of Optimized Problem Solving, which can be unfolded in four layers: Firstly, being cognizant of the dynamics surrounding the problem; Secondly, being receptive to a multitude of perspectives; Thirdly, iterating through various components by breaking them down to their fundamental levels, and finally, establishing a compelling logical connection across all the identified components.
I joined Brillio as a Data Scientist two and a half years ago, in April 2020 (right when COVID struck), and during my journey here I have worked on & spearheaded several AI/ML engagements for various customers across different industry domains, and also developed in-house AI/ML capabilities.
Solving complex customer business problems and ideating solutions not only during RFP’s but also during engagements requires a problem-solving framework, which primarily begins with identifying the right problems to solve & then translating those business problems into AI/ML problems. This gets followed by a rigorous application of the golden circle of Why’s-What’s-How, which implicitly drives me to widen the horizons w.r.t different perspectives & approaches and that in-turn leads to robust AI/ML solution development. The experience that I have gotten while working at Brillio has allowed me to express & nurture this solution mindset which in-turn enabled us to bag a multitude of deals & seamlessly execute the delivery of the same thereafter, thereby generating significant business value!
The role of a data scientist is very similar to that of a detective. How do detectives actually work? They look for patterns! Eventually, they uncover a sequence of specific different leads, and they connect all of them together to solve a specific case. However, not every step of detectives is 100% deterministic & they speculate a lot based on evidence. That's how detectives work, and that's what data scientists do with the data while solving business problems through AI/ML models.
Problem-Solving through first principles thinking in Data Science is considered the holy grail for robust solution development & Data Science is a combination of art & science that primarily becomes probabilistic when it comes to AI/ML models. Since these AI/ML models are expected to be an abstract representation of reality, they are bound to have an inherent component of uncertainty attached to them during development, as well as post-deployment, which often gets in the way of generating tangible business value.
Therefore, there are two potential concrete ways to minimize these uncertainties in AI/ML Models and thereby generate tangible business value:
a) The development of a toolkit that can enable us to tackle the uncertainty-chink in the AI’s armor by making predictions more explainable and detecting & mitigating biases (at the data & algorithm level), especially while taking high-stake business decisions.
b) The development of an ecosystem of thorough & rigorous AI/ML experimentation that answers the questions of “How to iterate quickly & what happens if we fail”.
The above two concrete ways formed the rationale for envisaging and developing end-to-end value-centric Cloud-Agnostic DS capabilities: Responsible AI(a) & ModelOps(b) – which are easily integrable into any Cloud Platforms. The ModelOps capability encompasses the overall ecosystem within which the AI/ML models are developed, thereby enabling us to arrive at the right perspective(model) rapidly, reducing the overall cycle time of AI/ML models, making predictions unbiased & explainable (Responsible AI) and enhancing the AI/ML adoption rate across different personas of Sponsors, Developers, Regulators, and Consumers.
I researched, conceptualized, and led the developmental efforts for building both capabilities from scratch. Throughout my journey with Brillio, translating visions into reality, with the strategy of sustainable value creation, and iterative communication across different levels (from the technical teams to the internal stakeholders), represented a significant learning opportunity & the same was brought to life through seamless execution of series of Technical-Learning Sessions, GTM’s, Series of Demos and Enablement Sessions. Also, communication plays an important role in simplifying complex concepts & articulating solutions to the clients, more so when we are trying to win deals & engagements. These efforts eventually led to presenting the capabilities to external stakeholders & winning numerous Customer deals, and it continues to bring in more pipelines and get resounding positive feedback from all the customers.
However, one cannot expect AI/ML models to perform consistently over a period of time because data/model drift is a reality of life, and, if not monitored, AI/ML models will decay silently. That is where the third capability comes into the picture – Timeless ML: Intelligent Model Health Monitoring, a novel value-centric capability that I ideated, designed, and conceptualized.
The Timeless ML: Intelligent Model Health Monitoring capability embeds Intelligence into the AI/ML Systems (for various classes of predictive problems) to detect data/model drifts in an automated manner even without the availability of ground truth, initiates automated Re-training mechanisms, and develops a robust causal framework for the quantification of the business impact of ML models post-deployment. This capability development involved extensive research, study & deconstruction of research papers, and then devising novel solutions, guided by the first principles of Data Science thinking. This Intelligent Model Health Monitoring capability enforces trust, enables reliable ML model performance/outputs, minimizes manual intervention, and improves the ROI of the deployed inventory of ML Models at scale, thereby generating much-needed business value.
An instrumental factor that enabled and empowered me to deliver such impactful outcomes is the rich & diverse culture of Brillio - from peers to senior management, ideas are welcomed & encouraged by providing a platform for experimentation. Team members provide different perspectives, and constructive feedback and get involved in the solution development, which, in turn, enabled me to polish & refine the ART of Optimized Problem Solving.