The Odos Project

The Odos Project is an employee-led institution that hires people to share their lived experience in lightweight, structured ways and turns that input into aggregated insights for organizations.

Comment

Inspiration

Odos started from a persistent frustration: stability is treated like a privilege. Things that should be basic like healthcare, benefits, predictable income, and safety often depend on the job you can access, not the life you’re living or the value you create.

I wanted to explore this problem by building a company that can materially support people with the least leverage while also creating undeniable value for customers. The core belief behind Odos is simple: people shouldn’t have their lives controlled by access to benefits and stability and society loses when we ignore the lived experience of the people who keep everything running.

That led to the Phase 0 framing: a worker-led, labor-backed insight engine. Workers contribute short, real check-ins about their lives and work. The platform turns that into aggregated, privacy-preserving insights that can help organizations make better decisions without turning individuals into products.

Tools like Gemini make this approach practical, they can take messy, human check-ins across voice, text and video and consistently turn them into summaries and structured fields in seconds. That lets contributors speak naturally while the platform still produces analysis-ready, aggregation-first insights without needing a huge team of human analysts.

What it does

I built an end-to-end prototype that demonstrates the full loop from worker signal → structured data → aggregated insight along with client tools like survey creation. Worker check-ins can include voice, text or video giving a mobile-first flow that makes it easy to share a quick update after a shift. Each check-in is transformed into a clean summary plus normalized categories/tags powered by Gemini so the data can be aggregated and analyzed not just stored as raw text.

The my log portal gives a transparent view where contributors can review what they submitted and how it was interpreted. The Client portal delivers a lightweight interface for partners to set up prompts/surveys and view results at an aggregate level. Plus a Gemini powered survey creator to help clients maximize the potential of the Odos network.

Finally, the public ledger shows an aggregation-first view that visualizes trends and distributions (e.g., category breakdowns, recurring themes), reinforcing accountability and trust.

How I built it

  • Gemini prompting: designed for “extract → normalize → summarize,” producing consistent structured outputs from messy real-world language.
  • Aggregation/analytics: rollups and visualizations that prove the platform can turn individual check-ins into cohort-level patterns.
    • Gemini powered analytics, voice to text capabilities, survey and power analysis, and generating an AI-driven network report of previous submissions to identify sentiment and trends across the network

Challenges I ran into

  • Signal vs noise: real check-ins include emotion, partial thoughts, and multiple topics so extracting structure without flattening truth is hard.
  • Trust-by-design: building something that feels respectful and transparent while still being useful to clients.

What I learned

  • The magic isn’t just transcription, it’s structured extraction and normalization powered by AI across a large evolving dataset that makes lived experience usable as aggregated insight.
  • Transparency matters: a worker-facing view of “what I contributed” helps this feel worker-led instead of extractive.
  • Designing for evolving question sets is critical, which means storing responses in a flexible but analyzable way to unlock iteration.
  • The hardest work long-term isn’t only technical, it’s creating a credible and ethical growth loop that earns trust on both sides of the market.

Built With

Share this project:

Updates