Soirée Démos IA

Trevor Longino|Makegreen.ai

Makegreen.ai is a free, open-source framework and engineering methodology developed by Green AI Studio to help companies drastically reduce their AI inference costs and energy consumption. Presented by founder Trevor Longino, the project addresses the "McLaren Problem"—the common, wasteful practice of running repeatable, predictable tasks on expensive, high-power frontier AI models when simpler, more efficient alternatives are available. The methodology is built around six core principles: defaulting to determinism (converting tasks to code instead of model calls), using the minimum sufficient model, auditing workflows before automating, measuring inference cost as a first-class metric, prioritizing local-first hardware where possible, and designing for auditability. To implement this, Makegreen.ai suggests installing a router that evaluates query complexity, routes tasks to simpler models (such as Qwen or DeepSeek), and only "fails upwards" to a frontier model if the simpler model cannot solve the problem. Longino shared real-world case studies demonstrating the framework's impact. For instance, greenchemistry.ai reduced its cost per process analysis from $5.00 to under $0.005 per run by converting 90% of its workflow to deterministic Python. Similarly, an AI copywriting detection platform cut its costs from $0.50 per million words to less than $0.003 by replacing stochastic LLM calls with a deterministic lookup table. To get started, users can visit makegreen.ai to access the open-source principles, playbooks, and fork

Neil Pitman|Social Media Assistant

Leah Mazur|Lift Curious

Lift Curious, founded by Leah Mazur, is a community platform designed to help women in Montreal find weightlifting partners, share knowledge, and build confidence in the gym. The project addresses a common gap for women who want to start strength training but face barriers such as a lack of guidance, unfamiliarity with programming, or intimidation in male-dominated weight sections. By connecting local gym-goers, the platform aims to make weightlifting more accessible and less intimidating. Using AI-assisted development, Mazur built and launched the fully functional minimum viable product in one week with no technical background, investing approximately 25 hours and 78 dollars in credits. On the platform, users create a profile detailing their local gym, experience level, fitness goals, and movement interests. Members can then browse a community directory to find recommended gym buddies with shared interests, send connection requests, view contact information once connected, and interact on a centralized community feed. During its early validation phase, Lift Curious ran targeted advertisements that generated 41 signups with a 90 percent profile completion rate. Moving forward, the platform plans to officially open its community directory on July 1, 2026, and host monthly in-person events, including learn to lift classes and local meetups. Women in Montreal looking for accountability partners and gym buddies are encouraged to join the waitlist at liftcurious.com.

Dan Le Van & Yann Jeudy|LifeMemory

LifeMemory is a voice-first storybook platform designed to help people preserve their life stories, targeting aging populations, immigrant families, and others whose histories are often lost. Users record memories by speaking into the app or uploading existing audio. The platform's AI analyzes these spoken memories, detects recurring themes, and links mentioned individuals to build a connected story graph.