AVM Generator Platform for Hepic

I led the development of the AVM Generator platform for Hepic (hepicmarketing.com), working together with Pro Internacional (prointernacional.com) and Delfos (delfoslabs.com), both part of the Domus Global ecosystem. The goal was to take Hepic's AVM model —their internal framework for planning content and campaigns— and turn it into a usable product instead of a manual, spreadsheet-based process.

Hepic is a B2B content marketing and demand generation studio that creates content in multiple digital formats (blogs, ebooks, scripts, podcasts, social, email, and more) to help organizations grow through content. Their AVM (Audience/Value Model) breaks down potential audiences for a product or service in a cascading way: macro segment → segment → interests and pains → media they consume → content formats that create impact. This model is powerful, but maintaining it manually in Excel meant weeks of research and iteration for each new client or campaign.

We transformed that manual process into a web platform built with Next.js on the front end and AWS Lambda functions as a serverless backend. At the core, we designed a wizard-style form that captures all the client's core information (industry, product or service, target audiences, objectives, tone, constraints) in a structured way, so the AVM logic can be automated instead of reconstructed from scratch every time.

On top of this, we integrated LLMs (OpenAI) to generate the full AVM matrix in minutes. From the initial brief, the system produces structured outputs for:

• Macro segments that could be interested in the product or service.
• Sub-segments with more specific needs and contexts.
• Interests, pains and motivations of each segment.
• Channels and media each segment is likely to consume.
• Content formats and angles with the highest potential impact for each segment.

The result is a cascading audience map that replicates Hepic's methodology but now generated programmatically. The platform outputs the AVM as structured data that Hepic's team can review, adjust and export, instead of starting from a blank spreadsheet every time.

From a technical standpoint, I designed and implemented the front-end architecture in Next.js, the serverless logic on AWS Lambda, and the LLM prompt and orchestration layer around OpenAI. We focused on:

• Keeping the form UX simple enough for non-technical users, while capturing all the context needed by the model.
• Structuring prompts and outputs so they map directly to Hepic's AVM dimensions (segments, interests, channels, content types).
• Ensuring the system can evolve with new prompt strategies, fields or export formats without rewriting the core.

Before this project, building one AVM required several weeks of manual research and spreadsheet work. With the AVM Generator, the Hepic team can create AVM matrices in minutes, and then invest their time where they add the most value: refining the strategy, validating insights with clients, and producing high-impact content, instead of wrestling with Excel.