Aman Tanwar
Portrait of Aman Tanwar

Hi, I'm Aman Tanwar.AI & Data Engineer

I turn scattered marketing data into systems teams can trust — and AI that answers their questions.

I spent my first five years inside Google and Microsoft's ad platforms, where I saw how often the numbers didn't agree. Now I build the pipelines, warehouses and AI agents that make them agree — and put them to work.

Get in touchSee my work
11+
Years in ad tech
70+
Data pipelines in production
99%
Cut in data tooling cost
Years of hands-on experience
  • SQL & databases
    8 yrsSQL & databases2018 – presentFrom troubleshooting with database queries to modelling data in BigQuery.
  • Ad platforms & performance
    5 yrsAd platforms & performance2015 – 2020Google Ads, SA360, Microsoft Ads — platform support and campaign management.
  • Tracking & analytics
    5 yrsTracking & analytics2021 – presentGTM, GA4, server-side tracking, consent and data collection strategy.
  • Web & platform development
    4 yrsWeb & platform development2022 – presentLed development of an ad tech platform with an offshore team.
  • Google Cloud
    4 yrsGoogle Cloud2022 – presentCompute Engine, App Engine, Cloud Run, Cloud Scheduler, BigQuery.
  • Data engineering
    2 yrsData engineering2024 – present70+ ETL pipelines, dbt medallion warehouses, identity resolution and attribution.
  • AI agents
    2 yrsAI agents2024 – presentText-to-SQL analysts and multi-agent systems on Vertex AI, Gemini and Google ADK.

My journey

Each skill built on the one before — from running ads, to measuring them, to engineering the platforms, data and AI behind them.

  • Ad platforms & performance2015 – 2020
    Ad platforms & performance2015 – 2020 · 5 yrsGoogle Ads, SA360, Microsoft Ads — platform support and campaign management.
  • SQL & databases2018 – present
    SQL & databases2018 – present · 8 yrsFrom troubleshooting with database queries to modelling data in BigQuery.
  • Tracking & analytics2021 – present
    Tracking & analytics2021 – present · 5 yrsGTM, GA4, server-side tracking, consent and data collection strategy.
  • Web & platform development2022 – present
    Web & platform development2022 – present · 4 yrsLed development of an ad tech platform with an offshore team.
  • Google Cloud2022 – present
    Google Cloud2022 – present · 4 yrsCompute Engine, App Engine, Cloud Run, Cloud Scheduler, BigQuery.
  • Data engineering2024 – present
    Data engineering2024 – present · 2 yrs70+ ETL pipelines, dbt medallion warehouses, identity resolution and attribution.
  • AI agents2024 – present
    AI agents2024 – present · 2 yrsText-to-SQL analysts and multi-agent systems on Vertex AI, Gemini and Google ADK.
View as table
AreaWhenYears
Ad platforms & performance2015 – 20205
SQL & databases2018 – present8
Tracking & analytics2021 – present5
Web & platform development2022 – present4
Google Cloud2022 – present4
Data engineering2024 – present2
AI agents2024 – present2

About me

I began in 2015 on the ads side: running Google Ads campaigns, then supporting Search Ads 360, Google Ads and Microsoft Advertising for Google and Microsoft through Accenture, Cognizant and HCL. The same problem kept coming up — clicks in one tool, conversions in another, revenue somewhere else, and nobody sure which number to believe. At HCL I started querying databases to find out why, and never really stopped.

In 2019 I moved to Canada, and in 2020 I joined an ad tech company. I started on measurement: fixing tracking, standardising the company's GA4 migrations, and launching server-side tracking and BigQuery onboarding as new services for 30+ clients. Each fix pushed me further downstream, into where the data lands and how it's used.

Since 2024 I've led data engineering: 70+ pipelines feeding a warehouse that powers reporting for 70+ clients, a platform that follows a customer from their first anonymous visit to closed revenue, and AI agents that let marketers ask that data questions in plain English.

Outside of work I build projects for the community — like an open pipeline for Vancouver's crime data, and OneTripDocs, which helps people get their OCI paperwork right the first time.

AI

  • Google ADK
  • Vertex AI
  • Gemini
  • LangChain
  • Multi-agent systems
  • Text-to-SQL

Data

  • BigQuery
  • dbt
  • SQL
  • Python
  • Medallion architecture
  • Identity resolution
  • Attribution

Google Cloud

  • Cloud Run
  • Cloud Functions
  • Cloud Scheduler
  • Cloud Build
  • IAM
  • Secret Manager

Ads & Analytics

  • Google Ads
  • SA360
  • Microsoft Ads
  • GTM
  • GA4
  • Server-side tracking

Projects

Real problems, what I built, and what changed because of it.

At work

Data engineering · Cost

Replacing a $20K-a-year Data Connector

The problem
We paid a third-party tool about $20,000 a year just to copy Google Ads data into BigQuery — and had to live with its schema.
What I built
An in-house pipeline on Cloud Run and Cloud Scheduler that pulls Google Ads data straight into BigQuery, in tables shaped for how we actually report.
The result
Tooling cost fell from ~$20,000 to ~$72 a year (99% less), and we own the data model.
PythonCloud RunCloud SchedulerBigQuery
Data platform · dbt

Visit-to-Revenue Attribution Platform

The problem
Our largest client, a destination resort, could see ad clicks and website visits, but not which of them turned into actual sales.
What I built
A multi-tenant bronze → silver → gold warehouse in BigQuery and dbt. It matches hashed GA4 user IDs to CRM customer IDs across 14,000+ sales, joins Google, Meta and DV360 ads, and is guarded by ~47 dbt tests, IAM deny policies between tenants, and CI/CD on Cloud Build.
The result
For the first time, the client can follow a customer from their first anonymous visit to closed revenue, and give each channel fair credit with multi-touch attribution.
BigQuerydbtGA4Cloud BuildIAM
See the public template →
Tech lead · Vertex AI

Ad Tech Platform & AI Analyst

The problem
Campaign teams juggled manual tools, and every client question meant someone pulling numbers by hand.
What I built
Led an in-house ad tech platform on Google Cloud with an offshore dev team, then added an AI analyst to it: marketers ask a question in plain English, and Gemini on Vertex AI writes the SQL and answers from the warehouse.
The result
Campaign teams spend 18% less time on analysis, because clients and marketers can answer their own questions.
Vertex AIGeminiLangChainBigQueryCloud Run
AI agents · Google ADK

Multi-Agent Data Analyst

The problem
Hard questions span CRM and ad data at once, and a single AI prompt writing SQL gets them wrong too often — or runs expensive queries.
What I built
Five agents in a fixed pipeline: one understands the question, one finds the tables, two write SQL in parallel, and a validator checks and retries (up to 3 times) before a final agent writes the answer. A cost circuit-breaker dry-runs every query against a daily spend cap.
The result
Checked, plain-English answers — with cheap models for routine steps, a stronger one only for the final write-up, and no runaway BigQuery bills.
Google ADKGeminiClaudeBigQuery
View on GitHub →
Data engineering · RevOps

One Company ID Across the Business

The problem
The same client had different names in the CRM, accounting and project tools. Invoices didn't match projects, and every dashboard and folder was set up by hand.
What I built
Pipelines (Python, n8n, BigQuery) that link HubSpot, QuickBooks, Monday.com and Teamwork into one hierarchy: company → domains → invoices → projects → campaigns, all keyed on a single company ID.
The result
Finance, marketing and campaign teams finally see the same client — and project managers' workload dropped 15%.
Pythonn8nBigQueryEntity resolution
Monitoring · Alerting

Ad Spend Anomaly Monitor

The problem
One wrong budget setting can burn through a client's money in hours, and nobody notices until the spend is gone.
What I built
A monitor that checks spend across Google Ads, Meta and DV360 every 15 minutes and alerts the team in Google Chat the moment something looks off.
The result
Budget mistakes are caught within 15 minutes instead of after the money is spent.
PythonCloud RunBigQueryGoogle Chat

For the community

Open data · Data engineering

Vancouver Crime Data Pipeline

The problem
Vancouver's public crime data comes as one huge raw file that's hard to analyse or map.
What I built
A pipeline on Google Cloud that loads the raw data, cleans it, and adds location features in bronze → silver → gold layers in BigQuery.
The result
A clean, analysis-ready dataset anyone can use for maps, research or machine learning.
BigQueryCloud StoragePythonSQL
View on GitHub →
AI product · In progress

OneTripDocs

The problem
People applying for an OCI card often get turned away or charged extra at the counter because of small mistakes in their documents.
What I built
A tool that reads your documents with AI, checks them against the official rules, and explains any problems in plain language before you go.
The result
Goal: one trip to the counter, no surprise fees.
Next.jsFastAPILLM document extractionPython
onetripdocs.com →

Experience

  1. Ad tech company · Canada

    2020 — Present

    Lead Systems & Data Engineer

    2024 — Present
    • Built 70+ ETL pipelines (Google Ads, SA360, DV360, Meta, GA4, HubSpot) feeding a warehouse that powers reporting for 70+ clients.
    • Architected a multi-tenant medallion platform on BigQuery and dbt for our largest client, with identity resolution and multi-touch attribution.
    • Replaced a ~$20K/year third-party connector with an in-house pipeline costing ~$72/year.
    • Led the AI layer: a text-to-SQL analyst on Vertex AI (18% less analysis work) and a multi-agent system on Google ADK.
    • Unified finance, CRM and project data under one company ID, cutting project managers' workload by 15%.
    • Mentor engineers on data modelling and dbt.

    Analytics Implementation Engineer

    2022 — 2024
    • Standardised the company's GA4 migration framework and trained new implementation engineers on it.
    • Launched server-side tracking, BigQuery onboarding and cookie consent as new services, onboarding 30+ clients.
    • Helped design the in-house ad tech platform, then led its development with an offshore team on Google Cloud.
    • Designed data collection and validation frameworks so data arrives in BigQuery clean.

    Technical Marketing Engineer

    2020 — 2022
    • Managed paid media campaigns in my first year, then moved fully into the technical side.
    • Built technical solutions across DV360, Campaign Manager, Google Ads and analytics, including a weather-data integration for contextual ad targeting.
    • Spotted measurement gaps and helped set up the company's technical services team.

  2. HCL Technologies · Microsoft project

    2018 — 2019

    Microsoft Advertising (Bing Ads) Support

    • Troubleshot Bing Ads delivery issues for advertisers, using database queries to find root causes.
    • Built automated diagnostic tools that made delivery troubleshooting faster and less error-prone.

  3. Cognizant · Google project

    2017 — 2018

    Google Ads Support Specialist

    • Supported high-priority Google Ads advertisers with account and campaign issues.
    • Solved attribution, tracking and data-quality problems, lifting conversion performance 18% across priority accounts.

  4. Accenture · Google project

    2016 — 2017

    Search Ads 360 Support

    • Supported agencies and advertisers using Google's Search Ads 360 platform.
    • Ran tagging QA and resolved data discrepancies between platforms — my first taste of data quality work.

  5. Prodigitas · Digital agency

    2015 — 2016

    Google Ads Specialist

    • Started my career planning and running Google Ads campaigns.

Let's talk

Hiring, a project idea, or just want to connect? Send me a message and I'll reply within a couple of days.

LinkedIn →GitHub →