Gaetano Zampogna
Arts, Digital Archives, AI
2026
London, UK
Gaetano Zampogna Artist Archive
A painter’s fifty-year catalogue, rebuilt as an artist archive visitors can explore and its artist can maintain.
A painter's fifty-year catalogue rebuilt as a public archive that visitors explore through a curator agent, and that the artist maintains through a private studio agent by conversation. The migration preserved 314 items and every old address, 1,222 checked live on 4 October 2026, while adding clearly labelled draft texts and the artist's approval for every public change.

My decisions, bets and trade-offs
- Decision
I chose an approval card and typed confirmation over one-tap publishing because a proposed correction was not the same as an approved change, and Gaetano needed to decide what reached the archive. Why
- Decision
I chose retrieval before writing stories because chapter-by-chapter generation failed in 9 of 10 runs, while retrieval-first failed in none of 16. Why
- Bet
I relied on the matcher to link candidate material to catalogue works, then checked its results against hand-matched works to see whether its suggestions held up. Why
- Trade-off
I accepted extra friction from the approval card and typed confirmation because a proposed correction was not the same as an approved change to an artist’s public record. Why
- 314Catalogue items migrated and checkedPreserved and checked the full catalogue during the archive migration.
- 186Missing work texts draftedAdded draft texts for works that previously had no text, helping fill catalogue gaps.
- $8.45Cost to generate final text setCreated the final set of missing-work texts at a total cost of $8.45.
- 1,210Verified legacy redirectsVerified old addresses still lead to the correct page or picture, preserving access to the archive.
Context
My father, Gaetano Zampogna, has painted for five decades. I replaced his WordPress site with an artist archive that lets visitors explore the work through a public curator and gives Gaetano a private conversational studio for maintaining it.
I designed and built the archive end to end: the migration, agent experiences, approval workflow, research and writing systems, and launch. That describes the scope of my contribution, not a different role title.
Challenge
The old site depended on a content management system my father was not going to maintain himself. Catalogue gaps therefore stayed open, and each addition depended on someone else. The rebuild also had to preserve existing writing and old links, distinguish drafted text from the artist’s own words, and keep unreviewed changes out of the public archive.
Approach
I migrated the catalogue while preserving its existing prose, then checked the old addresses against the live site. To research missing details, the studio agent reads family material from a shared Google Drive and scanned catalogues. OCR makes the scans searchable, followed by page-by-page review. A matcher links candidate material to catalogue works; I checked it against 164 works matched by hand. One register keeps open questions from being asked twice or lost, while a completeness score shows which gaps remain.
For works without text, I rebuilt the writing register after the first pass sounded wrong, then judged the rewrites against my father’s own writing. The resulting texts are labelled as drafts, not attributed to him.
Decisions
I chose an approval card followed by a typed confirmation phrase, rather than a one-tap publish action. That extra friction is deliberate: a proposed correction is not the same as an approved change to an artist’s public record. Gaetano remains the person who decides what reaches the archive.
naming convention nd image analisys, find related content or new content not listed and checked by coverage
Results
I migrated and checked all 314 catalogue items, confirmed that all 1,222 old addresses still lead to the right page or picture (tested live on 4 October 2026), and drafted texts for the 186 works that had none, at a cost of $8.45 for the final set.
Reflection
The low cost of generating text was not the hard part. The harder job was making its authorship clear and keeping the artist in control of publication. I also learned to review streamed answers as they arrive: a response can be correct at the end and still be confusing on the way there.
Prototype
Two agents share one catalogue but have distinct roles. The public curator answers visitors only from archive material and displays the works it refers to; the private studio converses with Gaetano and proposes new works, corrections, or missing details. An approval card separates proposals from publication: the artist must approve the card and type a confirmation phrase before a change reaches the public site.
Tools and Skills
End-to-end design and development of the archive.
Content strategy and editorial writing for the catalogue and the archive.
Sources from a shared Google Drive and scanned catalogues; optical character recognition (OCR) made the scans searchable, followed by page-by-page review.
The matching process was checked against 164 works matched by hand.
Agent design and operational planning for maintaining the archive.
Technology and Operations
Google Cloud Run behind a Cloudflare Worker; daily spending ceilings for the agents; analytics behind consent; every change to the archive passes an approval card.
Ux Accessibility
- Every screen is checked at widths from 320 to 1,280 px.
- Motion follows the visitor’s reduced-motion setting: fades remain, while other motion is removed.
- Pictures carry alt text in Italian and English.
- The site follows the visitor’s language, including technique names.
Experimentation and Rollout
I tested two ways of building the archive’s stories. Generating chapter by chapter failed in 9 of 10 runs. Fetching the relevant works first, then writing, failed in none of 16 runs. I chose the retrieval-first workflow based on those tests.
Disclaimers
Confidentiality
Details have been omitted where company policy, confidentiality agreements or legal obligations require it. What remains is accurate.
Fair dealing and my own views
Material is shown for portfolio purposes under fair dealing. No confidential or proprietary information from Gaetano Zampogna is disclosed, and the views here are mine rather than Gaetano Zampogna's.
Copyright, Designs and Patents Act 1988 ss.29-30; Data Protection Act 2018; Computer Misuse Act 1990.
Key facts
- Client
- Gaetano Zampogna
- Year
- 2026
- Role
- AI Design Engineer
- Industry
- Arts, Digital Archives, AI
- Location
- London, UK
- Catalogue items migrated and checked
- 314
- Missing work texts drafted
- 186
- Cost to generate final text set
- $8.45
- Verified legacy redirects
- 1,210
What was the challenge in Gaetano Zampogna Artist Archive?
The old site depended on a content management system my father was not going to maintain himself. Catalogue gaps therefore stayed open, and each addition depended on someone else. The rebuild also had to preserve existing writing and old links, distinguish drafted text from the artist’s own words, and keep unreviewed changes out of the public archive.
What did I actually do on Gaetano Zampogna Artist Archive?
I migrated the catalogue while preserving its existing prose, then checked the old addresses against the live site. To research missing details, the studio agent reads family material from a shared Google Drive and scanned catalogues. OCR makes the scans searchable, followed by page-by-page review. A matcher links candidate material to catalogue works; I checked it against 164 works matched by hand.
What was the outcome of Gaetano Zampogna Artist Archive?
I migrated and checked all 314 catalogue items, confirmed that all 1,222 old addresses still lead to the right page or picture (tested live on 4 October 2026), and drafted texts for the 186 works that had none, at a cost of $8.45 for the final set. Measured outcomes: Catalogue items migrated and checked: 314; Missing work texts drafted: 186; Cost to generate final text set: $8.45; Verified legacy redirects: 1,210.