Case studies and engineering notes.
Detailed write-ups of systems we have designed and shipped, and the engineering decisions behind them.
Systems we have built, written up in full.
Each study covers the problem, the architecture we chose, the trade-offs we made, and the engineering constraints that shaped the result. We write these from systems we have built and run ourselves, so the numbers in them are measured rather than illustrative.
Turning subjective description into reliable identification
A pattern-recognition engine that converts unstructured human description into a numeric attribute vector and matches it against a structured catalogue, entirely on-device. Plus the recommendation, profiling, and cost-controlled model integration built on top of it — shipped in ScentID, our own consumer product, and written up with the real thresholds and the real failures.
Written from work, not from theory.
Our notes cover the decisions that do not show up in a feature list: how we bound the cost of a model call, why a recommendation that cannot explain itself is worse than none, and what it takes to make an interface readable before anyone has to ask.
Recent posts on engineering and AI.
Why On-Device Inference Changes the Economics of AI
Three Independent Cost Ceilings: A Better Model for AI Spend
When a Test Cannot Fail, It Manufactures Confidence
Building Recommendations That Admit What They Don't Know
Entitlements That Don't Leak: Security at the Boundary
Release Engineering for Mobile: Beyond Feature-Complete
The next generation of software will be intelligent.
ScentID Solutions is building toward that future — one platform, one integration, and one intelligent solution at a time.
Work With Us