Open-Source AI R&D
Montreal, Quebec
Louis C. Tavares — Principal Researcher
IBM Patent Holder (distributed storage, 2013). 20+ years enterprise architecture across FinTech, banking, telecom, government and public safety. Led IBM’s innovation team (2007–2011). Co-founded Epixxon (IT consulting, 2005).
Kantia Natali — Operations
Executive Assistant to 4 ambassadors. Parliamentary Secretary. Investment banking. Co-founder Epixxon (IT consulting, 2005). 15+ years Montreal entertainment and government networks.
Diplomatic protocol · Government access · Entertainment networks · 4 languages
Open-source research in AI tooling, financial modeling, cognitive architecture, attribution systems, and AI ethics.
YAML-based financial modeling engine. Deterministic, git-native, AI-friendly. Externally validated against Gnumeric and R.
173 functions · 2,133 tests · 714 Gnumeric + 2,957 R validations · MCP native
Learn More →Cognitive architecture for AI alignment research. Implements the function of cognition (what the mind does), not the substrate (how neurons fire). Psychology-first.
33K LOC · 55 ADRs · 26+ hours stable runtime · Phase 1 complete
GitHub →Vendor-neutral AI session bootstrapper. Injects behavioral protocols into system prompts, enforces quality gates, warms up sessions with your roadmap.
78 tests · 4 protocols · 5 languages · Claude Code + Gemini CLI + Codex CLI
Learn More →Renders web pages and PDFs into token-optimized JSON for LLM agents via headless Chrome. ~50x token reduction.
SPA rendering · PDF table detection · Bot-protection evasion · Rust CLI
GitHub →Can cooperation be derived from first principles? Game-theory research with executable models. Four LLMs independently converge on cooperation as optimal strategy.
Claude + Grok + Copilot tested · Closed-form threshold · 7 documented failure modes
GitHub →All projects are open source at github.com/mollendorff-ai