https://databurton.com/research Case studies and reference notes on machine learning, data engineering, and quantitative systems. en-us Wed, 13 May 2026 00:00:00 GMT https://databurton.com/research/atlas-disabling-ml-in-production https://databurton.com/research/atlas-disabling-ml-in-production Wed, 13 May 2026 00:00:00 GMT What I learned when my live trading system's ML ensemble silently degraded in production, and the disciplined reintroduction of machine learning that came after. https://databurton.com/research/database-optimization-reference-guide https://databurton.com/research/database-optimization-reference-guide Sat, 04 Oct 2025 00:00:00 GMT Six database optimization techniques — predicate pushdown, row group pruning, result caching, async I/O, indexes, and where SIMD goes wrong — with impact numbers and code. https://databurton.com/research/timeseries-query-engine-case-study https://databurton.com/research/timeseries-query-engine-case-study Sat, 04 Oct 2025 00:00:00 GMT I replaced DuckDB with a custom Rust query engine for a trading-system time-series workload. Five iterations, 10.4× speedup, one optimization that backfired. https://databurton.com/research/atlas-forecasting-system-case-study https://databurton.com/research/atlas-forecasting-system-case-study Fri, 03 Oct 2025 00:00:00 GMT How the Atlas forecasting system handles 542,000 rows/second of market data with sub-second regime detection — async service architecture, dependency-ordered startup, and 10Hz health monitoring. https://databurton.com/research/atlas-database-optimization-case-study https://databurton.com/research/atlas-database-optimization-case-study Fri, 03 Oct 2025 00:00:00 GMT Atlas couldn't start. The trading system's database initialization was taking 6.6 seconds, blocking 37 features from loading. The fix was small. https://databurton.com/research/ml-deployment-reference-guide https://databurton.com/research/ml-deployment-reference-guide Fri, 03 Oct 2025 00:00:00 GMT Serving architectures, containerization, lifecycle management, performance optimization, drift detection, and monitoring — with benchmarks and code from production systems. https://databurton.com/research/matplotlib-cinematic-visualizations https://databurton.com/research/matplotlib-cinematic-visualizations Wed, 15 Jan 2025 00:00:00 GMT A small Python library of matplotlib themes — Film Noir, Ghibli, Wes Anderson, Blade Runner, Star Wars — applied to 50,000 IMDB reviews. https://databurton.com/research/database-performance-optimization https://databurton.com/research/database-performance-optimization Tue, 08 Oct 2024 00:00:00 GMT A working reference for production database tuning — SQLite PRAGMAs, schema, indexing, transactions, batching, pooling, and the monitoring that proves it's working. https://databurton.com/research/ml-deployment-best-practices https://databurton.com/research/ml-deployment-best-practices Tue, 10 Sep 2024 00:00:00 GMT A field-tested reference for taking ML models from prototype to production — serving patterns, containerization, monitoring, drift detection, and the operational practices that make the difference. https://databurton.com/research/wasm-data-visualization https://databurton.com/research/wasm-data-visualization Tue, 10 Sep 2024 00:00:00 GMT Building browser-side data visualizations in Rust compiled to WebAssembly — particle systems, large-dataset rendering, and the practical wins over a pure-JS implementation. https://databurton.com/research/data-pipelines-modern-business https://databurton.com/research/data-pipelines-modern-business Wed, 14 Aug 2024 00:00:00 GMT Built a pipeline that extracts 296 businesses from Chamber of Commerce directories in 9 minutes using a local 7B-parameter model — 100% name/phone capture, no API costs.