Applied AI · Product experiment

AI-Assisted Workspace

A production pilot with model-agnostic AI and responsive long-running workflows.

Joseph's role
Tech Lead
Experience
Work led by our founder
Documented outcome

~2 days

to test models instead of multiple weeks

A model-agnostic layer made comparisons faster.

Scope of the work

  1. Product discovery

  2. Model portability

  3. Queued AI workflows

The challenge

An exploratory workspace needed to turn multi-step AI generation into a usable product experience. Model experiments took weeks, while individual workflows could take minutes on the models available at the time.

Joseph's contribution

Product discovery

Turned the initial idea into an MVP scope, release roadmap, risk register, and a set of assumptions the client could test.

Model-agnostic integration

Designed and implemented the AI layer so the team could compare models without rebuilding the product around one provider.

Asynchronous execution

Moved long-running AI work to queued cloud workers and defined the surrounding infrastructure, secrets, and deployment pipeline in Terraform.

What the work enabled

The pilot shipped to production, reduced model experimentation from multiple weeks to a few days, and gave the client evidence for the product's next direction.

Working through a similar problem?

Tell us what your team needs to build or improve. We'll help identify a useful next step.

Book an intro call

Explore more work

All projects

Cloud operations · Technical audit

Production Performance Audit

A focused audit connecting production symptoms to infrastructure and application bottlenecks.

View project

Enterprise software · Compliance

Security & Compliance SaaS

Frontend foundations that supported a growing team, followed by 20+ backend integrations across regulated communications platforms.

View project