BurlyMingo Labs
Current products, research, infrastructure, and open technical or creative work from BurlyMingo.
A wholesale restoration lab for photographers, archivists, and independent capture businesses. Partners capture slides and positive film for their own customers, retain the originals and customer relationship, and submit supported source files to BurlyMingo for source-preserving restoration.
Active product and primary commercial focusOngoing work improving large-scale restoration of aging photographic film, including color recovery, white-balance selection, collection consistency, damage handling, negative-film conversion, and clearly separated optional AI enhancements.
Active research and production developmentA practical benchmarking platform for testing local and frontier language and vision models against real agentic tasks, tools, prompts, hardware, and deployment constraints. The work identifies where smaller local models are genuinely useful, where frontier models remain necessary, and where a task exceeds the reliable capabilities of any current model.
Active developmentBurlyMingo’s self-hosted environment for model inference, workflow automation, image processing, and agent experimentation across workstation GPUs and Jetson hardware. It includes dynamic model routing, local OpenAI-compatible endpoints, Hermes, Ollama, image-generation workflows, and resource-aware orchestration.
Active infrastructureA local-first orchestration and policy layer for connecting AI models with tools, routing requests, and controlling how capabilities are exposed.
Experimental developmentWorldbuilding and narrative-development experiments focused on maintaining consistency across geography, factions, characters, history, rules, and evolving fictional settings.
Creative researchA LitRPG- and tabletop-inspired fictional universe shaped by old-school MUD culture, collaborative storytelling, and assisted worldbuilding tools.
Long-term creative projectHeritage is the commercial priority. The surrounding research and infrastructure support restoration work, local model evaluation, and longer-horizon experiments without being presented as customer-ready products.