Profoundry

Profoundry is a private internal Streamlit app I built to make a best-in-class GEO platform much easier to use at agency speed.

This is an intentionally high-level write-up. I work with Profound data in a professional context, so I am not sharing client information, internal reporting details, or anything that would turn this page into an HR meeting with dramatic lighting.

What I can share is the pattern, and I think the pattern matters.

The problem

Profound is extremely strong as a GEO tool. That was never the issue.

The issue was operational friction.

When you need data across multiple topics, multiple clients, and multiple pulls, manual exports stop feeling charming very quickly. The UI had export options, but the workflow was still too click-heavy and too repetitive for the volume we needed. Good data was sitting behind a process that took hours when it should have taken minutes.

That is exactly the kind of problem vibe coding is absurdly good at solving.

What Profoundry does

Profoundry gives me a simple internal interface for handling the boring part:

• connect to the Profound API
• pull the needed data across topics and clients far faster
• normalize the outputs into a repeatable flow
• store the results in our data lake instead of relying on one tool UI forever

The win was not making something flashy. The win was collapsing a manual process into something faster, more consistent, and easier to reuse.

Why this project clicked for me

This was one of the clearest moments where vibe coding stopped feeling like “fun internet product experiments” and started feeling like a real professional advantage.

I did not need to start a new business. I did not need to ship a public SaaS. I needed a better workflow.

Codex and Claude helped me move from “this process is annoying” to “this process now has software” without turning it into a months-long internal tools saga.

That matters because a lot of valuable software should never become a startup. Some of it should just quietly save smart people a huge amount of time.

The stack

• UI: Streamlit
• Language: Python
• Source data: Profound API
• Storage: data lake ingestion pipeline
• AI build partners: Claude Code and OpenAI Codex

The technical choice was mostly about speed and usefulness. Streamlit let me stand up an internal UI quickly, Python handled the API and data work cleanly, and the AI tooling helped accelerate the glue code, iteration, and edge-case cleanup.

What makes this portfolio-worthy even though it is private

I am not linking to a live version because it requires authentication and touches client data. That is not negotiable.

But private does not mean unimportant.

If anything, this is a better example of what practical AI-assisted building looks like in a real company environment:

• identify a workflow that is expensive in human time
• build guardrails around the implementation
• keep sensitive details out of the public story
• still ship something that materially improves the work

That is the kind of innovation I want to keep doing more of.

What companies should take from this

The lesson here is not “everyone should build a GEO export helper.”

The lesson is that teams are full of slow, annoying, high-value processes that never get fixed because they do not seem big enough for a formal product roadmap and too technical for a non-engineer to tackle alone.

Vibe coding changes that math.

If you understand the workflow, can describe the pain clearly, and can test the result with some discipline, you can build useful internal software much faster than most organizations are used to.