.png)
.svg)
.svg)
.svg)
.svg)
In a time when business is being restructured at unprecedented speed due to AI, it’s easy to forget that the fundamentals of data analysis — clear problem definition, grounded statistical reasoning, and business context — haven’t gone anywhere. Generative AI and large language models are transforming what's possible, but they don't eliminate the need to understand what you’re solving for in the first place.
Ashish Mohan knows this balance well. A seasoned data science leader, Ashish has worked across insurance, retail, infrastructure, and urban planning. He blends deep technical knowledge with practical business acumen, and in our conversation, he shared what more data professionals and business leaders need to hear: Don’t start with the tools — start with the problem.
“The core of all machine learning models is just statistics,” he said. “Until you know your business problem, you won’t know what analytical method to use.”
Ashish’s insight is not a critique of AI. In fact, he sees AI as a powerful tool — but one that must be matched to a clearly articulated need. When business teams reach for LLMs before defining the core challenge, they risk building architectures that are impressive, but ultimately misaligned.
In our recent conversation, Ashish offered a rare combination: a cross-industry lens and a grounded, ROI-driven approach to data science. We covered what he sees as today’s biggest pitfalls, how to deliver meaningful insights, and the hidden edge of generalists in a specialist world.
Across verticals, Ashish sees business leaders pushing for AI-driven solutions without a clear understanding of the problem they’re trying to solve.
“Sometimes people default to generative AI. But what they need could be solved by regression models,” he said. “You don’t need ChatGPT to predict sales. A good linear model will do.”
He emphasized that LLMs and generative AI tools have incredible potential, especially when used to improve data accessibility and speed up exploratory work. But jumping to AI without grounding the business question risks misalignment and wasted effort.
While some leaders recommend specializing in one vertical, Ashish sees value in moving between them. From spatial modeling in urban planning to marketing analytics in insurance, he’s transferred mental models across industries.
“Customer segmentation in urban planning isn’t that different from consumer segmentation in retail,” he said. “The concepts are portable. The labels just change.”
His ability to connect concepts across domains has given him an edge in spotting opportunities others miss. Ashish believes that exposure to different industries strengthens a data scientist's creativity, adaptability, and business judgment.
“Gone are the days where you could work in a single vertical and know it inside out. Today, people interact with data across multiple platforms and touchpoints. The ability to think cross-functionally is critical.”
Ashish doesn’t romanticize the work. He sees data leaders spending too much time doing what he calls “data plumbing”: de-duplicating campaigns, standardizing offline media, and manually making taxonomy decisions. That isn’t avoidable — but it needs to be done smartly.
“You don’t need to process 100% of the data,” he said. “Start with a question. What stream of the data do you actually need?”
He recommends actively de-scoping models and working closely with clients to limit noise before handing anything off to engineering teams.
He also noted that tools like Excel, Python, and SQL Server — used well — are often sufficient for both internal analysis and presentation. While cloud-native stacks are powerful, business users often gravitate toward familiar, interpretable tools.
.png)
Ashish emphasized the importance of midstream client check-ins. Most data scientists engage clients at the start and the end of a project. But critical value is created in between.
“Why wait until the final model to talk to them?” he asked. “Early results can help you course-correct — and help the client rethink their assumptions.”
In his view, presenting the insights well matters just as much as building a model that works. That means delivering context, not just numbers.
“You can make a fancy dashboard with heatmaps and bubbles,” he said. “But if a simple bar chart tells the story, use that.”
Asked what advice he’d give to someone starting in the field, Ashish didn’t hesitate:
“Don’t start with the tools. Start with the statistics,” he said. “Then learn how it applies to business. Tools come and go. But the principles stay.”
He sees too many new practitioners rushing into Python packages or LLM integrations before mastering the underlying math. Concepts like regression, multicollinearity, and PCA remain the foundation — no matter how advanced the tooling becomes.
“If you don’t understand R-squared or MAPE or the reason a variable matters, the numbers are just numbers. They won’t tell a story.”
It’s a reminder for all of us: the foundation of great data science isn’t the model. It’s the clarity behind why we’re building it in the first place.
.png)
The streaming tv platform Tubi was launched in 2014. Its revenue had reached $150 million by 2019, and by 2023 the revenue was $900 million.
Read Post
In our latest installment of the "Data Leaders Uncovered" series, we sat down with Maxwell Dylla, a data scientist with a background in materials science engineering.
Read Post.png)
We’re excited to continue our "Data Leaders Uncovered" series by sharing insights from Prem K. Narasimhan, a seasoned executive in the global life sciences industry. Prem has been at the forefront of addressing some of the most complex challenges in pharmaceutical research, clinical trials, and data management. This conversation shines a spotlight on how innovation and strategic collaboration can drive meaningful change in the industry.
Read Post
Welcome to the second installment of our blog series, "Data Leaders Uncovered." In this series, we delve into insightful conversations with data leaders across industries, exploring their challenges, innovations, and lessons learned in navigating the ever-evolving data landscape. In this post, we continue our discussions by spotlighting Jared, A Director at W.E. Aubuchon Company; a hardware retailer based in New England.. We discussed the intricacies of managing retail data, the transition to modern tools like Snowflake and Power BI, and the opportunities presented by AI-driven solutions.
Read Post
We’re excited to start our new blog series dedicated to conversations with data leaders. Each post will explore their challenges, innovations, and lessons learned in navigating the ever-evolving data landscape. We will be highlighting challenges they have overcome, pain points and the achievements they take the most pride in.
Read Post
CRM systems have long served as the primary hub for customer data. Platforms like Salesforce and HubSpot excel at tracking customer interactions, managing sales pipelines, and providing basic reporting. These functions have made CRMs indispensable for sales and marketing teams seeking a centralized view of customer relationships.
Read Post
In recent years, the concept of fractional professionals has gained significant traction, especially in tech companies. Fractional executives, such as fractional CFOs, CMOs, and RevOps, provide their expertise to organizations on a part-time or project basis, offering a cost-effective and flexible solution for companies that may not require or cannot afford full-time executives in these roles.
Read Post.png)
A Go-To-Market (GTM) organization may find it beneficial to invest in a Data Warehouse (DWH) alongside their Salesforce CRM when they encounter specific indications that their data and analytics requirements have surpassed the analytics and reporting capabilities of Salesforce.
Read Post.png)
Data quality is a critical aspect of any data warehouse project. With an ever increasing number of data sources and technologies, scoping and prioritization is daunting. In this document, we share our data quality framework for planning data quality actions, current state assessments, and prioritizing data quality initiatives.
Read Post.jpeg)
In modern data pipelines, continuous data integration and data frequent updates may cause unintended data errors. Blue-green Deployment ensures a final checkpoint for accurate results. In this blog, details are provided on how to implement blue green deployment in a data warehouse project.
Read Post
Companies targeting mobile app publishers can use mobile app data to identify potential customers. This blog discusses using bulk mobile app data for sales outreach.
Read Post.png)
Discover how applied AI can automate tasks and enhance productivity in data analysis. Explore a real life example of Applied AI in report generation.
Read PostStay in the loop with everything you need to know.
%203-svg%20(2).png)
