Product strategy, UX/UI design and AI-assisted development
of an agentic content research platform for pharma marketing

EN DE

My role on the project

CommsIQ began as an internal initiative: AI-driven research that keeps pharma marketing teams in the loop on the topics they care about, so they stay relevant and can amplify the ones that matter to them.

I shaped the product strategy, designed the UX/UI and built the first MVP myself with AI-assisted development, on Google Cloud and its Agent Development Kit. Product decisions, interface and code all sat with one person.

Along the way I plugged in several third-party generative AI tools, for video, visuals and more, and tested how each fits a team’s content workflow.

Problems I solved

A pharma team knows more than it ever says in public. Its posts come in bursts: a campaign, a reshare, then silence. The people who could speak with authority are its biggest untapped channel, and they don’t have the time to read everything and write about it.

So the product turns that work into one steady loop, the one the animation above plays. Content tracks set the narratives, audience and cadence. Research agents watch topics, trends, people and signals while the team gets on with its day. Every finding is categorised and scored, so people choose what deserves their opinion. AI-assisted workflows turn it into a draft in their own tone of voice, ready to schedule on a shared calendar.

The goal is to move a team from random social media activity to consistent narrative delivery that builds trust.

Experience I can bring to your project

  • Agentic, multi-agent AI products
  • AI-assisted development, from concept to live
  • Development with Google Agent Development Kit (Google ADK)
  • Product strategy for an MVP
  • Integrating third-party generative AI for video and visuals
Comms studio

Define topics, audience, strategy and enable content tracks.

Research agent 128 signals found

Creative AI radar

Research agent 74 signals found

Pharma launch watch

Research agent 96 signals found

Regulated trend scan

Signals

Findings are categorised and prioritised. Inspire or express expert opinion.

Personalisation

→ Research preferences
→ Tone of voice
→ GenAI visual styles

Content workflows

→ Signal opinions
→ AI-assisted editor
→ Professional services

Autoposting

→ Direct posting
→ Schedules
→ Team calendar

Campaign Videos Posts Podcasts Intra
One loop: content tracks, research agents, signals, then posts for every channel.
Set up an agent. Users create research agents in an AI chat: describe the topic, confirm what it picked up, and the research starts right away.
Work with signals. Users explore the important signals their research brings in. AI categorises each one by the criteria they set.
Create a post. A signal becomes a draft in the user’s voice: rewrite it, add a visual, preview it on LinkedIn and schedule it.
Work in Studio. Teams follow their content tracks and people, then plan a new track and its owners in the same AI chat.
Multiple agents run in parallel. Procedural graphics show each one’s research at work.
Comms team

Researchers and editors, in the browser.

Firebase Auth

Google sign-in, an email allowlist, session cookies.

LinkedIn OAuth

Each person connects their own account to post.

Console · Cloud Run

Next.js 16 and React 19: studio, agents, signals and posts, with its own API.

Cloud Scheduler

Wakes the engine every five minutes.

Gemini API

Drafts, rewrites, proofreading and post images.

Cloud SQL

Postgres 16 with row-level security: agents, findings, posts.

LinkedIn API

Publishes the posts, with their images.

Secret Manager

Keys and database credentials.

Engine · Cloud Run

Python and Google ADK: discover, collect, analyse and score findings.

Vertex AI

Gemini 2.5 Pro plans the search, Gemini 2.5 Flash reads each finding.

Google Search

Grounding for discovery, with Custom Search as fallback.

Web sources

RSS, Reddit, GitHub, LinkedIn and pages.

A content research, posting and distribution platform: two Cloud Run services around one Cloud SQL database, agents on Google ADK and Gemini. I designed, built, tested and ran it on this architecture myself.
Next case Glacier Express