Project PresoGen AI
Commercial AI Desktop Application — Multi-Provider LLM Pipeline
The Challenge
Technical presenters waste hours building slides. SaaS presentation tools require subscriptions, upload content to third-party clouds, and offer no awareness of the project being presented. The presentation workflow had not been reimagined through AI-native thinking.
What I Built / Delivered
Built a local-first desktop application (Electron 41, React 19, TypeScript) with a mandatory plan-before-generate pipeline — a 7-state session machine requiring human approval before generation. Multi-provider LLM support (OpenAI, Anthropic, Google) with structured JSON output (Zod schemas). Workspace-aware context indexing. Tri-format export (HTML, PDF, PPTX). Commercial licensing, encrypted API key storage, ~600 automated tests.
The Outcome
Feature-complete product in pre-launch stage. Demonstrates structured LLM pipeline engineering, secure local-first architecture (BYOK, no cloud upload), and full product thinking from engineering through to pricing and go-to-market.
Why This Matters
A practitioner who builds commercial AI tools — with production-grade testing, structured validation, and a real business model — demonstrates AI Mindset at the product level. This is not a consulting deliverable; it is evidence that AI-native thinking produces shippable products.