Research & Novel Systems
Sleep-Cycle Continuous Learning in LLMs
An LLM architecture that implements biologically-inspired sleep cycles for continuous learning. The system separates experience accumulation (wake) from weight consolidation (sleep), allowing the model to learn from direct user interaction without catastrophic forgetting.
Drawing on memory consolidation literature, including complementary learning systems theory and hippocampal replay. The sleep phase replays and distills new experiences into stable long-term representations. End-to-end implementation: data pipeline, training loop modifications, evaluation harness.
Exploring applications in personalization, domain adaptation, and lifelong learning for deployed models.
Articles
Research
Engineering & Systems Building
12 years of production software
Founded and ran Tipasoft, a development agency delivering web, mobile, and desktop applications across a dozen industries. Full-stack engineering: architected systems from scratch and integrated into complex existing codebases. Comfortable across frontend, backend, mobile, and infrastructure.
Select clients
Cross-Domain Pattern Recognition
Why breadth matters for AI research
The pattern of entering unfamiliar domains, building accurate mental models quickly, and shipping. It's the same skill continuous learning systems need to acquire.
Music Production
Co-wrote, recorded, mixed, and mastered electronic music. Worked with Grammy-winning producers. Managed live performances at Canada's largest venue (Rogers Centre). Nominated for Toronto Independent Music Awards. Trained intuition for signal processing, temporal patterns, and iterative creative refinement.
Consumer Products & Education
Created a coffee brand and children's robotics kits. Teaches kids robotics. Experience designing physical and digital systems, and thinking about how non-experts learn.
Enterprise Across Industries
Clients across finance, automotive, CPG, media, and education. Each required rapidly building domain models and translating between technical and non-technical contexts.