Gaurav Tiwari

Data & AI leadership

Data your people, and your AI agents, can trust.

AI is only as good as the data it reads and the checks on what it changes. I get both ready: the foundation agents can use, and the gates that keep their changes safe. Then I help hire the team that owns it.

Book a free 30-minute call See how I help

Engineering Manager at Spotify, leading the verification of AI-generated code and data changes. Previously data leadership at Meta, home24 and Lazada.

Selected clients

01Where you are

Where is your data today?

  1. Stage 1

    Scattered

    Data lives in many tools and nobody is sure which numbers to trust.

    • Teams quote different numbers for the same thing
    • Reports are built by hand, every time

  2. Stage 2

    Understood

    You know what exists and what is broken, but there is no foundation to build on.

    • Each new question means a new one-off query
    • Legacy tools hold the business logic

  3. Stage 3

    Built

    A platform is in place, but nobody senior owns the roadmap or the team.

    • You are about to make your first data hires
    • Data work competes with everything else for attention

  4. Stage 4

    Trusted

    People rely on the numbers, and AI is starting to change the code that produces them.

    • AI tools are writing or reviewing pipeline changes
    • You need to show what changed, and why

Where this ends

Agent-ready

Your people and your AI agents work from the same trusted definitions, and every AI change passes the same gates as a human one.

02What I do

Four ways to get your data ready for AI, from a short diagnostic to ongoing leadership.

  1. AI-readiness diagnostic

    2–4 weeks

    Can your data support AI agents, and what has to change first?

    I map your sources, definitions and reports, find where the numbers stop agreeing and what an agent would get wrong, and give you a prioritised plan for what to fix first.

  2. Safe AI in the data workflow

    Scoped to your team

    Let AI change your pipelines without losing trust in the numbers.

    CI/CD for data, quality gates and verification for AI-assisted pipeline changes, so every change is checked against thresholds and leaves an audit trail.

  3. Agent-ready data foundation

    6–12 weeks

    One set of trusted definitions for your people and your AI tools.

    Hands-on build or migration: a warehouse, a tested dbt project (staging, intermediate, marts), a semantic or metrics layer, and reporting in Looker or Metabase.

  4. Fractional Head of Data & AI

    1–2 days a week

    Data and AI leadership before you are ready to hire it full-time.

    Strategy and roadmap, advising the CEO or CPO on where AI fits, and hiring your first data people, with someone senior who can still write the models.

The AI work comes from my day job: at Spotify I lead the verification of AI-generated changes, including data pipeline changes, with quality gates, thresholds and an audit trail.

03How it works

Built to end with someone else owning the work.

  1. Diagnose

    Talk to leadership and the people who use the data. Map what exists, where it lives, and where the numbers stop agreeing.

    You leave withA clear picture of your data and a prioritised roadmap.

  2. Build the foundation

    Hands-on: warehouse, tested dbt models and a BI layer, with quality checks in the delivery pipeline from the start.

    You leave withA platform your team can extend, not one only I understand.

  3. Set up the leadership

    Shape the data function and its roadmap with the CEO or CPO, and help hire the people who will own it.

    You leave withA team that owns the work after I step back.

Not every company needs all three. A diagnostic can stand on its own, or you can bring me in for just the build or just the leadership.

04Selected work

Problems I have been brought in to fix.

Advisory engagements alongside my full-time roles. Clients are described rather than named.

Challenger bank

Hands-on, alongside the backend engineering team

  • Metabase
Problem
Reporting was archaic and disconnected. The databases existed, but were poorly understood.
What I did
  • Diagnosed what data existed and where it lived.
  • Set up Metabase as the core reporting tool.
  • Worked with the backend engineers to build the pipeline from source systems to a reporting layer.
Outcome
Reporting infrastructure built from near zero.

Series B media platform

Hands-on across data engineering, product analytics and analytics engineering

  • BigQuery
  • dbt
  • Looker
Problem
Analytics ran on a legacy SAS platform, and the product team needed analytics it could work with.
What I did
  • Worked as data engineer, product analyst and analytics engineer.
  • Built product analytics dashboards, then helped design a new data platform architecture.
  • Migrated the SAS platform to BigQuery and dbt, with Looker for consumption, and owned the full dbt layer: staging, intermediate and marts.
Outcome
The legacy SAS platform replaced by BigQuery, dbt and Looker, with one analytics engineering layer behind the reporting.

AI inference hosting platform

Advising the CPO, and hands-on with the data

  • Stitch
  • Segment
  • BigQuery
Problem
Go-to-market and advertising data was fragmented across tools, and there was no defined data function.
What I did
  • Unified the go-to-market and advertising data across Stitch, Segment and BigQuery, with dashboards on top.
  • Worked with the CPO to define the data function, a two-year roadmap and the key hires.
Outcome
A clear plan for the data function, which led to a conversation about joining full-time as Head of Data.

M&A marketplace

Six months, hands-on, working with the CEO

Problem
Sellers were not matching with the right buyers.
What I did
  • Ran a six-month, hands-on engagement diagnosing why matches were going wrong.
Outcome
Helped the CEO hire a Director of Analytics to own the roadmap.

I have also worked on engagements in agriculture, e-commerce, tourism and weather data.

05Background

From founding a data function to leading engineering at scale.

  1. 2021 – present

    Engineering Manager, Spotify

    London

    • Leads engineering teams across AI testing (verifying AI-generated changes, including dbt in a data monorepo), GenAI prototyping, and Apollo, the Java framework behind thousands of microservices used by 900+ engineers.
    • Built fleet-wide quality monitoring and a verification framework with merge-quality thresholds, calibrated separately for backend, frontend and data code, and an auditable feedback store.
    • Partnered with an external research team on a study of 72,215 pull requests, which found that component-level and integration test coverage predicts post-merge defects.
    • Previously led Soundtrap analytics and growth: an 11% signup uplift, and a CMS rollout that lifted signups by 8% and organic traffic by 12%.

  2. 2019 – 2021

    Data Analytics Manager, Meta

    Dublin

    • Led 10 analysts across three time zones.
    • Owned the analytics roadmap, standards and governance, forecasting and lifecycle analytics.

  3. 2015 – 2019

    Data Architect & BI Lead, home24

    Berlin

    • Founded the data function from zero: the team, the roadmap, the first data warehouse and the ingestion framework.
    • Introduced self-serve BI (Tableau) across the company. Spoke at conferences.

  4. 2014 – 2015

    BI Lead, Lazada

    Kuala Lumpur

    • Led a team of 10 analysts.
    • Put machine-learning pricing and fraud detection models into production, alongside forecasting.

Education

  • M.S. Applied Information SystemsSingapore Management University, 2014
  • B.E. Electrical EngineeringRGPV, India, 2012

Tools and skills

  • dbt
  • BigQuery
  • Snowflake
  • Databricks
  • Airflow
  • SQL (advanced)
  • Python (advanced)
  • Java
  • Looker
  • Tableau
  • Metabase
  • Hex
  • CI/CD for data
  • Data quality and observability
  • AI-assisted code and data review

06Speaking

On stage, and one to one.

  1. 2024 – present

    Industry speaker, BrainStation

    London

    Industry speaker panels, including "Building a career in data".

  2. 2015 – 2019

    Conference speaker, while leading data at home24

    Berlin

    Spoke at industry conferences while founding the data function.

Mentoring

  • MentorCruiseMentor
  • Data Action Mentor1:1 mentoring for startups on data, analytics and AI, since 2024
  • iO SphereData mentor, London, since 2024

Topics I speak on

  • Verifying AI-generated code and data changes
  • Getting your data ready for AI agents
  • Building a data function from zero
  • Building a career in data

Running an event or a team offsite? Get in touch.

07Writing

Notes on the work, coming soon.

  • In progress

    Verifying AI-generated changes

    Quality gates, thresholds and audit trails for letting AI change code and data pipelines safely.

  • In progress

    Building a data function from zero

    What to build first, who to hire first, and when to hand it over.

08Contact

Tell me where your data is today and what you need it to do.

Start with a free 30-minute call, and we can work out whether a diagnostic, a build or ongoing leadership fits. Or message me on LinkedIn or by email.

Book a free 30-minute call