White Matter Podcast

Dashboards used to be where decisions started. At Opendoor, 80% of decision workflows began in one. Today they begin in AI tools, and the data team's job has changed with them.

Rahul sits down with Paras Doshi, former Head of Data at Opendoor and early Alexa analytics lead at Amazon, for a candid conversation about what happens when decision makers stop opening dashboards and start asking AI instead.

If you lead a data team, build analytics infrastructure, or want to stay relevant as AI agents become the first stop for every business question, this episode is for you.

Paras explains why the most-used feature in almost every dashboard tool has always been "export to Excel", and why AI has quietly replaced that step. When the data, not the dashboard, becomes the product, the people who define metrics, build semantic models, and own verification loops end up in the critical path.

He also breaks down how to decide where AI should run freely and where humans must stay in the loop, using a simple framework of two-way door and one-way door decisions, along with a real example of how one bad call on a single home can wipe out the profit on a hundred others.

You'll discover:

Why decision makers at Opendoor moved from dashboards to AI tools

How data engineers are becoming owners of metric definitions and semantic models

The three A's every data professional needs now: adaptability, agency, and ambition

Where AI should decide, and where tail-risk decisions still need a human

How to think about build vs buy with a "Lego architecture" for AI-native data stacks

Why Opendoor built its own AI-native data catalog when no vendor had caught up

Paras also shares why early adopters of AI get a compounding edge that late movers struggle to close, why he is working more hours but has never been happier at work, and what new grads should do right now in a job market that is harder than ever. His advice is direct: find what you can offer that the current workforce can't, then go build real systems to prove it.

An AI system can be a little wrong when it drafts an email. It cannot be a little wrong when its answer may affect someone's credit.

Rahul Pattamatta speaks with Swamy Ramajayam, Chief AI and Data Officer at Mariner Finance, about what companies need before AI can move from answering questions to influencing real decisions.

They discuss why customer-facing AI still needs human validation, why the dashboard is no longer the real product, why curated data and shared definitions matter, and how the analyst's role changes when AI can produce the first answer.

Swamy brings more than 18 years of experience across Meta, Ally Financial, Charles Schwab, Capital One and IBM.

White Matter is presented by DataBrain.

Dr. Suman Bhattacharya, Senior Director of Data Science and AI at Docusign, joins Rahul Pattamatta on White Matter.

They discuss why AI evals are not new, where human guardrails still matter, how AI is changing knowledge work, and why software built for people cannot simply be repurposed for agents.

Suman also demonstrates intentLM, his open-source work for turning customer behaviour into structured intent that agents can use.

Contribute to intentLM:
https://www.intentlm.ai/contribute

Data can now answer questions in plain language, write queries and find patterns. But should it also tell a business what to do?

Rahul Pattamatta speaks with Shivram Lakshminarayanan, Head of Analytics at Swiggy, about where data should stop and human judgment should begin.

Shivram brings 16 years of experience across Mu Sigma, Classle, Brillio, PhonePe and Swiggy. At PhonePe, he rose to Head of Analytics & BI before taking on his current role.

They discuss why reading the chart is no longer enough, why data definitions matter when AI is doing the querying, the difference between knowing SQL and knowing where to use it, and how the analyst’s role moves closer to the business.

White Matter is presented by DataBrain.

The best data analysts are no longer just answering questions. They are designing the systems that decide which questions matter, how fast answers arrive, and whether any of it actually changes the business.

Rahul sits down with Prudhvi Vasa, who leads the data team at Postman, for a sharp conversation about how analytics work is evolving from one-off requests to action-oriented, feedback-driven decision-making.

If you lead a data team, hire analysts, or want to stay relevant in a world where LLMs and agents are changing the workflow, this episode is for you.

Prudhvi breaks down why so many analyses fail the most important test - “so what?” - and how strong teams prevent wasted effort by forcing every request to connect to a real business outcome.

He explains the shift from centralized data support to embedded models, when domain expertise matters, and why problem solving beats tool obsession every time.

You’ll discover:

• Why great data work should end with an action item, not just an insight
• How Postman thinks about hiring for analysts, data engineers, and embedded support
• Why tools like SQL, R, Python, and BI platforms matter less than systems thinking
• How data modeling is becoming more important as agents start querying data directly
• What it means to move “left” in the stack from reporting to modeling to infrastructure decisions

Prudhvi also shares how the analyst role is changing in the age of AI, why human judgment still matters, and what young professionals should do right now to stay useful as the space gets more automated. His advice is simple and practical: get your hands dirty, build real projects, and learn to talk about impact, not just effort.