AI Showcase recap header photo
MontrealNovember 3, 2026

Event Recap

Inside Montréal's AI Scene: Four Builders Present Hardware, Memory, Infrastructure, and Urban Planning AI

On November 3, 2026, 129 engineers, founders, and local tech enthusiasts gathered in Montréal for the latest edition of the AI Showcase. Across four concise pitches, presenters demonstrated practical deployments spanning humanoid robot cognitive architectures, urban traffic triage, hardware-optimized video generation, and deterministic multi-agent workflows. The room brought sharp technical focus paired with live software and hardware demos.

Inside Reachy: Giving a Humanoid Robot Graph RAG Memory and Ebbinghaus Forgetting Curves

Soich Inc. developed a persistent dual-memory Graph RAG system with Ebbinghaus forgetting curves for the Reachy Mini humanoid robot.

Damien Soichet of Soich Inc. introduced an architecture giving Pollen Robotics' Reachy Mini humanoid persistent long-term memory. By combining Neo4j graph databases with vector embeddings, the system maintains context across sessions using a dual-memory approach where System 2 consolidates memories and applies Ebbinghaus forgetting curves during idle states. The room got a live look at the setup in action, highlighted when Reachy spoke directly to the audience and joked about remembering embarrassing details about Damien.

Soich Inc

Damien Soichet presented the project for Soich Inc., addressing common memory loss and temporal limitations in LLM agents.

Cocoon Triage: Rerouting Traffic and Work Zones Across Montréal

Cocoon Triage converts construction site data into safe work-zone layouts and visual planning packages for traffic engineers.

Rashid Mushkani presented Cocoon Triage by Cocoon Lab, a tool aimed at helping municipal teams and traffic engineers design safer detours around urban construction sites. The platform transforms existing site parameters and plans into alternative work-zone layouts, complete with drawing packages and visual models. Cocoon Lab is currently putting the technology to work through an active pilot project with the City of Montréal.

Cocoon Lab — Cocoon Triage

Rashid Mushkani represents Cocoon Lab, which is actively testing street planning tools in a pilot project with the City of Montréal.

NPU-Driven Video Generation and Next-Gen AI Hardware

VDO-X leverages NPU hardware infrastructure to generate AI videos in 10 seconds while reducing compute latency.

Gabriel Salem, a computer engineering student at McGill University, presented VDO-X, an AI video generation platform designed around Neural Processing Unit (NPU) architecture. By executing inference calls on backend servers running distilled models on NPUs rather than relying on standard GPU bottlenecks, VDO-X optimizes speed and compute costs. Salem proved the infrastructure's performance live on stage, generating a complete AI video clip via a web interface in roughly 10 seconds.

VDO-X

Gabriel Salem is a computer engineering student at McGill University building application layers and frameworks for NPU hardware.

Avalanche: Turning Black-Box AI Agents into Industrial-Grade Workflows

Avalanche is an open-source framework by Trampoline AI that structures autonomous AI agent tasks into observable, sandboxed Python pipelines.

Émile Riberdy of Trampoline AI showcased Avalanche, an open-source control plane built to transition unpredictable AI agent setups into deterministic, code-defined Python pipelines. Each step within an Avalanche workflow runs in an isolated sandbox with durable checkpoints, quality gates, and runtime tracking to eliminate black-box execution failures. Riberdy demonstrated its real-world application by triggering an automated multi-agent code review pipeline directly from a live GitHub push.

Avalanche

Émile Riberdy presented Avalanche on behalf of Trampoline AI, creating deterministic control planes for enterprise task automation.

Connections and Conversations on the Floor

Once the final pitch concluded, the 129 attendees filled the floor to inspect the hardware up close and dissect implementation details with the presenters. Discussions ran late into the evening as traffic planners, developers, and robotics researchers exchanged ideas and explored potential collaborations across Montréal's growing AI ecosystem.
AI Showcase networking after the presentations

Want to be at the next one?

See upcoming events