AI Operations Lead

Why this role exists
Puretech runs creative, media and performance work at agency pace. Generative AI is already changing how that work gets produced — but adoption today is uneven, undocumented, and dependent on whoever happens to be curious. This role exists to turn that into a system.
You will make AI a dependable part of how campaigns get built: reusable prompt architectures instead of one-off experiments, and automated content pipelines instead of manual variant production. Equally important, you will own the analytics side — pulling campaign, content and AI-usage data from sources that do not agree with each other, reconciling them, and turning them into something a client or a pod lead can act on in thirty seconds. You will also be the person who teaches everyone else to use it well.
This is a hands-on role. You will build the workflows yourself, then hand them over.
What you will own
Prompt systems and quality standards
• Build reusable prompt architectures — system instructions, dynamic variables, structured output formats — for SEM ad groups, social scripts, long-form SEO articles and creative briefs.
• Configure Gemini Gems to client brand voices, personas, visual guidelines and negative-prompt rules, so output is on-brand without a human rewrite.
• Define quality rubrics covering fact-grounding, brand safety and advertising-platform compliance, and make them enforceable rather than aspirational.
• Maintain a shared internal prompt library with version history, so improvements compound instead of living in individual chat threads.
Content automation pipelines
• Build automated pipelines using Google’s generative media models — Imagen for visuals, Veo and Google Vids for video concepts — to produce ad variants, social banners and storyboards at volume.
• Enable creative and copy pods to work with Gemini directly inside Docs, Slides and Drive, rather than copy-pasting between tools.
• Deploy task-specific assistants and background agents to remove repetitive steps from campaign ideation, copywriting, asset generation and client review cycles.
Analytics, synthesis and reporting
This is half the role, not an afterthought to the AI work. We need someone genuinely good with data — not only someone who can configure a dashboard tool.
• Pull and combine data across GA4, Google Ads, Meta Ads, Search Console, and internal delivery and AI-usage data — sources with different definitions, granularities, attribution windows and refresh cadences.
• Reconcile them honestly. Where two platforms disagree on the same number, work out why, decide which one answers the question being asked, and say so rather than quietly picking the flattering figure.
• Write the underlying queries and data models yourself — SQL in BigQuery, or a well-structured Sheets model — rather than depending on someone else to prepare the data.
• Move from numbers to findings. Identify what actually changed, separate signal from normal variation, and resist reporting a correlation as a cause.
• Build and maintain real-time performance dashboards in Looker Studio that answer a specific question for a specific audience — not a wall of every available metric.
• Present the result so a non-technical reader gets it immediately: the right chart for the question, a clear headline finding, and the detail available underneath rather than in front.
• Report on campaign KPIs, content output velocity, and AI usage and cost — for internal reviews and client-facing reporting alike.
• Make reporting repeatable. Automate the refresh, document the definitions, and make sure a number means the same thing this month as it did last month.
Integration, cost and governance
• Connect Google Workspace to Vertex AI, BigQuery, webhooks and automation tools (Make or Zapier) so pipelines run without manual triggers.
• Track quota limits, cloud spend and per-campaign AI cost, and flag when a workflow stops being economical.
• Apply data-privacy and client-confidentiality standards consistently across accounts — particularly around what client data enters which tool.
Team enablement
• Train creative directors, copywriters, media buyers and SEO/SEM specialists on practical AI use — not tool demos, but the specific workflows their pod runs.
• Set the standard for when AI output is good enough to ship and when it needs a human pass.
• Measure adoption honestly, and fix the workflows people quietly abandon.
Success in the first 90 days Two production content pipelines live and in daily use by at least one pod. One dashboard, blending at least three data sources, that replaces a manual report and is trusted enough that people stop rebuilding it by hand. A documented prompt library with named owners. A baseline measurement of AI usage and cost across accounts, with a view on where it is worth spending more.
What we need from you
Must have
• Generative AI, hands-on: 2+ years building and shipping GenAI workflows in a delivery environment — prompt frameworks, assistants or agents that other people actually used. Personal experimentation alone will not meet this bar.
• Analytics — multi-source synthesis: Demonstrable experience combining data from three or more systems that do not naturally join, reconciling their differences, and producing one trustworthy view. This is a hard requirement, not a preference. Be ready to describe a case where two sources disagreed and how you resolved it.
• Analytics — from data to decision: You can move past describing what the numbers say to identifying what changed, why it plausibly changed, and what someone should do about it — while being honest about what the data cannot support.
• Data modelling and querying: Comfortable writing SQL against BigQuery, or building non-trivial Sheets data models. You should be able to design the underlying model, not only visualise someone else's.
• Visual communication: You choose a visualisation because it fits the question, and you can explain a finding to a creative director or a client without them needing to read the legend twice.
• Dashboards: Demonstrable Looker Studio work (Power BI acceptable if the underlying data modelling is strong). You should be able to walk us through a dashboard you built, why you structured it that way, and what you deliberately left out.
• Digital marketing literacy: Working fluency in GA4, Google Ads and Meta Ads reporting, and what SEO/SEM teams optimise for. You do not need to have run campaigns, but you must be able to hold a conversation with someone who does — including about why two platforms attribute the same conversion differently.
• Scripting: Python or JavaScript for API calls, automations and glue code.
• Google AI ecosystem: Practical exposure to Gemini for Workspace and Vertex AI — not just awareness of them.
• Location: Based in Kolkata, or able to relocate before joining. This role is on-site to start with – later could be hybrid.
Should have
• Hands-on use of Google’s generative media stack — Imagen, Veo, Google Vids, AI Studio.
• Agent orchestration experience, whether GCP-native (Agent Builder) or framework-based (LangChain, LlamaIndex).
• Experience automating workflows with Make, Zapier or equivalent.
• Comfort with basic statistical judgement — sample size, significance in A/B tests, and when a movement is just noise.
• A track record of getting non-technical colleagues to adopt something new.
• B.Tech / B.E. / B.S. in Computer Science, IT, Data Science or a related discipline — or equivalent demonstrated capability.
Nice to have
• Google Cloud certification (Associate Cloud Engineer, Professional ML Engineer, or Cloud Digital Leader).
• Experience connecting REST APIs into Looker Studio or Sheets via connectors or webhooks.
• Prior time inside a creative, advertising or performance marketing agency.
• Experience evaluating AI output quality systematically rather than by impression.

Work arrangement
No