Sunnyvale, California

Transforming AI ambition into measurable product impact.

I’m Sirshendu Maiti, a Senior Staff Data Science Manager at Google Search. I lead teams shaping the measurement, experimentation, and evaluation systems behind Agentic Search and AI Mode. We translate complex signals of quality, user value, and model performance into product strategy and high-stakes launch decisions across Search.

Previously, I held data science and analytics leadership roles at Meta, eBay, Barclays, and Citigroup.

Sirshendu Maiti
What I believe Agentic AI will redefine how people search, decide, and act online
What I lead Agentic Search · AI Evaluation · Measurement Strategy · Experimentation
What I Know Best
Search Metrics Experimentation Causal Inference Human & AI Evaluations Prompt Engineering Bayesian & Sequential Testing Statistics Time-Series Forecasting Risk Analytics Bayesian Decision Theory Network Effects & Interference
Thinking in Bets
My Book recommendation
Thinking in Bets by Annie Duke book cover

Annie Duke’s framework for making better decisions when outcomes are uncertain.

Thinking in Bets: Why Great Leaders Think in Probabilities

A short reflection on decision quality, uncertainty, and better leadership.

A lot of business leaders still think about strategy like it’s chess. If we gather enough data, analyze it hard enough, and make the right moves, we should be able to predict the outcome.

But business (and life!) is much closer to poker.

We’re constantly making decisions with incomplete information. Markets change, competitors react, customers behave in unexpected ways, and sometimes luck simply plays a role.

So a great decision can still lead to a bad outcome. And a bad decision can occasionally work out. That’s the central idea behind Annie Duke’s "Thinking in Bets".

What I like about the book is that it pushes leaders away from black-and-white thinking. Instead of asking, “Was I right or wrong?”, it encourages us to ask, “Given what I knew at the time, was this a good decision?”

It also teaches you to become more comfortable with uncertainty…shift from binary thinking to probabilistic reasoning…to say, “I’m 70% confident,” rather than pretending to be 100% sure… and then keep updating that view as new information comes in.

For statisticians: this is a sequential Bayesian learning process. You start with a prior P(H), update it with new evidence through P(D|H), and arrive at a posterior P(H|D). Today’s posterior becomes tomorrow’s prior.

And it gives you practical ways to challenge your own thinking, like red-teaming your strategy or doing a premortem before a major launch.

Leadership takeaway

Your job isn’t to win every bet. Your job is to build a high quality decision-making process that gives you better odds over time.

That’s why I think Thinking in Bets is such a valuable book for anyone making high-stakes decisions.

WHAT I LEAD NOW

Leading Data Science for Agentic Search & Real-World Journeys

Dedicated research, analytics, and evaluation organization supporting Google’s next-generation AI Search experiences for Google’s highest monetizable verticals, such as, Local, Travel and Finance. Our team empowers product, engineering, and UX leadership with actionable insights, innovative metric development, rigorous causal experimentation, and comprehensive human-and-AI evaluation systems.

IMPACT

Building the Systems Behind High-Stakes AI Decisions

Here are some of the initiatives (high level) I’m proud to have led with my teams — transforming ambitious AI ideas into trusted products and measurable impact.

**Views are my own and do not represent Google.

01

Research, Measurement & strategy

Defining how Agentic Search is measured and evaluated

Contributing to the development of next-generation Agentic experiences on Google Search

Leading Data Science strategy for emerging Agentic Search experiences , with a focus on understanding how AI-assisted Search helps people navigate complex, multi-step tasks and move from discovery to action.

Developed measurement and evaluation approaches that bring together experimentation, user behavior, satisfaction, and human-and-AI evaluation to understand product quality and user value.

Metrics Developed multidimensional approaches to evaluating AI-assisted Search experiences, helping teams understand quality across multiple aspects of the user experience.

Research Advanced research into how users perceive value in complex AI-assisted journeys and how successful task outcomes can be evaluated consistently.

Agentic AIMeasurement & Evaluation0-to-1 Product StrategyExperimentation
02

Experimentation & Metrics Research

Measurement for AI-Native Search

Developing measurement approaches for next-generation AI Search experiences

AI-assisted interfaces change how people interact with a product, which makes legacy engagement metrics difficult to apply directly - a measurement challenge the whole field is working through. I led Data Science work to build evaluation approaches suited to that shift, combining experimentation, behavioral signals, user feedback, and human evaluation. Partnered across Product, Engineering, UX Research, and evaluation teams to strengthen measurement rigor and support evidence-based product decisions at scale.

Key Impact Established more rigorous measurement foundations for next-generation AI Search, enabling better-informed product decisions.

Causal inferenceA/B testingDecision ScienceAI Search Metrics Innovation
03

AI Mode & Product Strategy

Advancing AI-Powered Experiences in Google Finance

Improving AI response quality and evaluation for financial research experiences

Led Data Science strategy for emerging AI-powered experiences in Google Finance helping shape how advanced AI capabilities support investors with research, analysis, and information discovery.

Key Impact Built evaluation frameworks for AI-generated financial analysis, supporting a rigorous, research-oriented experience for people analyzing markets.

AI ModeHuman EvaluationPrompt-Engineering

Leadership philosophy

Rigor builds confidence. Leadership creates momentum.

I trained as a statistician at the Indian Statistical Institute (Kolkata), where I learned to challenge assumptions, find signal in complexity, and make decisions grounded in evidence.

I’m a future-forward, results-driven leader and a committed coach to managers and technical leaders. I lead beyond the analysis—setting direction, aligning teams around consequential product outcomes, and building the measurement systems and leadership capacity required to deliver at scale.

My goal is simple: empower organizations that make better decisions, execute with velocity and conviction, and create lasting product impact.

AI & Data Science Leadership Organizational Leadership People Management Leadership Multiplication Statistical Rigor High-velocity Decisions
01

Set direction

Define the user problem, the decision, and what success means.

02

Create strategic alignment

Align data science roadmap directly with product goals— unite teams around shared outcomes, priorities, and trade-offs.

03

Build for scale

Turn insights into repeatable systems, operating mechanisms, and decision frameworks.

04

Multiply leadership

Develop managers and technical leaders who can drive outcomes independently.

05

Prioritize ruthlessly

Invest in the few initiatives that create the greatest customer and organizational leverage.

06

Stay agile

Move with speed, learn from evidence, and adapt the strategy based on user needs, technology shifts, and evolving market conditions.

Education

Indian Statistical Institute (ISI), Kolkata

One of India’s premier institutions, globally recognized for its contributions to statistics, quantitative research, and data science.

Master of Statistics | 2004–2006

Bachelor of Statistics | 2001–2004

QUANTITATIVE FOUNDATION

Regression & Machine Learning Multivariate Analysis Linear Algebra Probability Theory Statistical Decision Theory Optimization Research Time-Series Analysis Survival Analysis Stochastic Processes Game Theory Actuarial Modeling Experimental Design Statistical Inference

The Indian Statistical Institute (ISI) was founded in 1931 by Prasanta Chandra Mahalanobis, widely known as the "Father of Indian Statistics" for introducing the Mahalanobis Distance and pioneering large-scale sample surveys. Read his official biography here .

ACADEMIC DISTINCTION

Two-Time Indian National Mathematical Olympiad Finalist · 1997 & 1999

LinkedIn