Pixel Wisdom Start a project

Independent product engineering

Critical software, taken from hard problem to shipped product.

Philippe Tremblay designs and delivers complex web, desktop, data, integration, and AI products for technical teams—end to end.

Tell me what needs to ship See the work
  • Direct collaboration
  • End-to-end delivery
  • Québec, Canada

Proof

Work that shipped.

Selected products and engagements across learning, research interfaces, native software, and systems integration—owned from problem framing through delivery.

01 Product engineering · Data interfaces

ARC Prize Foundation

Owned end-to-end implementation of visualization and data interfaces for the ARC-AGI benchmark, working with the team and Mike Knoop.

Role
End-to-end product engineering
Scope
Task exploration & model evaluation
Output
Interactive benchmark UI
  • Visual task explorer for abstract benchmark puzzles
  • High-density comparison matrices and metadata architecture
  • Research surface for navigating tasks and model performance
02 Systems integration · Research workflow

Zotero ↔ NotebookLM

Built for Julia Turc, founder of Storia AI: a cross-app sync system that sends Zotero collections to NotebookLM and brings source guides, source content, and notes back into Zotero.

Architecture
Stateless polling bridge
Modes
Consumer bridge & Enterprise API
Output
Bidirectional research workflow
  • Collection sync with remote-first duplicate detection and binary PDF uploads
  • Reverse import of source guides, source content, and NotebookLM notes
  • Reverse-engineered NotebookLM’s undocumented RPC endpoints and payload formats
  • Authenticated browser bridge over a localhost task queue—without storing credentials
03 Independent product · Language learning

PhraseLang

Designed and built a sentence-first language-learning product that combines listening, recall, speaking, and spaced repetition in one responsive practice loop.

Role
Creator, designer & engineer
Stack
React, TypeScript, SpacetimeDB
Output
Responsive practice & reading app
Sentence practice Focused repetitions built around real language
Speech systems Recorded and on-device voices across languages
Synced progress Settings, level, and review history across sessions
View live demo
PhraseLang signed-in practice dashboard with new-sentence and review sessions
PhraseLang active English-to-German sentence repetition session
04 Independent product · Native macOS

Snappy

Designed, engineered, and launched a native file and app search product built to stay sharp on million-file home directories.

Role
Creator, designer & engineer
Stack
Rust, React, Tauri, SQLite
Output
Signed & notarized macOS app
200 ms Snapshot load to first search
3–5 s Full walk of ~1M entries
<250 ms Filesystem change to result

First-party measurements on a roughly one-million-entry test corpus.

View live product
Snappy macOS app search interface showing Terminal, Visual Studio Code, Notes, and Photos

Offer

One critical product outcome. Full ownership.

For technical teams that need a senior product engineer to hold product intent, architecture, and delivery—without another chain of handoffs.

  • Product framing Clarify the user, constraints, and smallest complete outcome.
  • Interface design Make complex states, data, and AI behavior understandable in use.
  • Full-stack delivery Own the surface from data model and APIs to the final UI.
  • Integrations & ship Join difficult systems, tune the critical path, deploy, and hand off.

About

Philippe Tremblay

Pixel Wisdom is my independent practice. I work where product decisions, architecture, and implementation meet—especially when a small team needs someone to hold the whole problem.

Public work spans research interfaces, browser and desktop software, data-heavy applications, local systems, and the details that make ambitious products usable by people who did not build them.

GitHub Pixl Studio

Inquiry

What needs to ship?

Share the product, the hard part, and where things stand. Philippe will review the brief and reply directly if the engagement looks like a good fit.

Selected interface work

ARC-AGI task results explorer with dataset and model filters above a dense evaluation table
01 / 06 Task results explorer A filterable research surface for inspecting task-level model evaluations.

Selected workflow

Zotero NotebookLM settings showing automatic consumer bridge setup and sync options
01 / 06 Integration configuration Consumer bridge setup, health checks, and sync behavior in one Zotero preferences surface.