Domains
Photobiology, tissue culture, low-power radio systems, process instrumentation, sensor fusion, contextual ML, and other domains where the useful answer is rarely available off the shelf.
Low-power hardware, analytics, machine learning, and durable systems.
Distributed sensing, geospatial analytics, contextual prediction, plant systems, and data infrastructure built to stay useful.
Current direction
Large-scale IoT analytics, sensor fusion, geospatial and spatiotemporal analysis, contextual prediction, and long-lived field hardware.
Good problems
Experimental programs that need to become disciplined systems without losing the lessons that made them interesting.
Applied work
Prototype to learn, operate to know, build with purpose.
Method
Reduce ambiguity, find the real operational constraint, and build the minimum system that can carry the idea into production.
Examples
Geospatial time series, contextual prediction, long-lived embedded sensing, production plant systems, and data architectures that can keep up with live analytical workloads.
Constraints
Portability, observability, operating cost, maintainability, and whether the resulting system remains legible after handoff.
Projects
A distributed sensing and analytics platform combining ultra-low-power hardware, geospatial context, and operational decision support.
A modular environmental control system for plant production, built around instrumentation, sensor fusion, and plant-response feedback.
Custom acclimatization infrastructure that turned a fragile tissue culture program into a usable production input.
Built an analytical chemistry lab and decision-support pipeline for evaluating plant material, metabolite composition, and production outcomes.
Principles
Directness, legibility, and sensible boundaries.
Stay close to the real problem
Most abstractions are useful until the work becomes real. When the edges show up, I prefer to step closer to the actual area of concern.
Use what holds up
New ideas matter, and so do systems that have already proved they can hold up. When a newer system is clearly better, it is worth using for that reason.
Keep the loop short
Self-hosting, bare metal, and declarative systems are useful when they shorten the feedback loop and keep the system coherent. Managed services are useful when they create real leverage.
Spend effort where it matters
Right-size the system for the objective, respect the real constraints, and put effort where it creates durable value.
Writing
A short note on the exchange year that pushed biotech and organic agriculture into the same frame.
A note on maintaining microbial libraries and testing formulations under real greenhouse and field constraints.
Just for fun
Shells, shaders, desktop tooling, and other small detours live mostly on GitHub.