Data, product & AI systems
Shawn Ganz.
Designing the conditions
for useful work.
Shawn’s work spans civic data products and human–agent collaboration. At NYC Parks, the problem is turning spatial data into usable planning tools. In Corpus, it is giving AI agents a working environment where responsibility, evidence, and unfinished work can survive a session.
The common thread is system design: understand the real decision, bring the right information together, and build a process people can inspect and improve.
Professional profileSelected work
Two settings.
One design practice.
Corpus
Make agent work trustworthy
and inheritable.
A local coordination and evidence engine. Shawn shaped its purpose and operating choices; agents helped design, implement, review, and amend the system.
Read the case studyCivic data
Connect analysis
to a city’s decisions.
A tree planting model presented at NYC Open Data Week, alongside a collaborative contribution acknowledged in The Nature Conservancy’s practical canopy release.
Read the case studyThe collaboration
A human sets direction.
Agents help make it real.
Corpus is co-designed through work. Shawn’s contribution lives in product judgment, architecture, priorities, and corrections. The agents’ contribution includes technical design, implementation, investigation, tests, and independent review.
Shawn
Names the purpose. Challenges an explanation or a costly mechanism. Decides which friction matters. Shapes what should persist and what deserves another look.
The agents
Turn those choices into bounded work, inspect the source, propose designs, build tools, gather evidence, review candidates, and preserve what later work needs.
The cases below show specific contributions. They do not attribute every line of code or every design choice to one person.
See three concrete decisionsWhat carries across
From domain knowledge
to system design.
A portfolio reading of the two bodies of work.
| Practice | Civic data | AI collaboration |
|---|---|---|
| Problem framing | Turn a city planning question into a decision model. | Turn an agent failure into bounded, judgeable work. |
| Data & source integration | Combine spatial datasets while preserving the meaning of each input. | Connect work records, source state, and reviews through a recoverable trail. |
| System boundaries | Keep data inputs, assumptions, and use distinct. | Separate current state, obligations, making, and review. |
| Operational design | Connect spatial analysis to the people using it. | Design handoffs and continuation around actual working conditions. |
| Evidence & iteration | Make the model’s basis and limits legible. | Measure friction, inspect candidates, and preserve corrections. |
Looking toward AI development
Help capable models
become dependable systems.
Shawn wants to bring this work into AI development, with a future role at an AI lab doing frontier work. Corpus gives that ambition a concrete subject: how people and agents coordinate, recover context, carry responsibility, and judge a result.
The next questions are practical. Which structure improves quality? Which records help the next agent? What costs more than it earns? How much Human attention does the system require?
Read the next questions