Research area 05
Data science and applied machine learning
The methodology of turning fragmented data into tested predictions, with baselines first, uncertainty shown, and every result reproducible from raw sources.
- methodology
- statistics
- baselines
- reproducibility
Applied machine learning lives or dies on method. The model is usually the easy part; the hard parts are the data lineage, the splits, the baselines, the leakage checks, and the honesty of the report. LAISER’s applied projects follow a single operating discipline, and we publish that discipline as a reusable framework so others can hold their own work to it.
What we work on
- Evidence labelling. Every claim in an analysis is marked verified, inferred, assumed, hypothesis, or unknown, and the label travels with the claim.
- Baseline-first modelling. Simple models, out-of-sample tests, and effect sizes before any complexity is added.
- Idea-to-analysis workflows. Tooling that lets a domain expert propose a question and get a rigorous, reviewable analysis back.
Projects in this area
Land Alpha
An early-stage research programme testing whether fragmented public and local data can identify land-market transitions before those changes are fully reflected in market signals.
- Spatial and geospatial intelligence
- Forecasting and decision intelligence
- Data science and applied machine learning
Principal Data Scientist Framework
A data-first operating system for ML and analytics work with AI coding agents, with evidence labels, leakage and point-in-time discipline, baseline-first modelling, adversarial review, and self-testing regression evals.
- Data science and applied machine learning
- Safety, alignment, and evaluation
- Multi-agent and agentic systems