The Applied AI Wiki is a set of reference articles on the machine-learning and data-science topics MLAIA works on in practice: audio and signal intelligence, medical AI, causal analysis, and large language models in production. Each entry is written as a neutral, encyclopedic overview — definitions, established methods, evaluation practice, and the trade-offs practitioners actually face — rather than as opinion or product material. Pages are reviewed by MLAIA's data-science team and carry an explicit last-updated date, so they can be read, linked, and cited with confidence.

The wiki complements the MLAIA blog: blog posts discuss timely industry developments and case studies, while wiki entries cover the underlying methods that change slowly. Where a wiki topic has a related blog post, the two link to each other.

Articles

  • Acoustic Event Detection and Classification

    How machines detect and label sounds — glass breaks, machine faults, sirens, speech activity — in continuous audio. Covers spectrogram and MFCC features, learned audio embeddings, hybrid DSP–ML detectors, the labeling problems of rare events, event-based evaluation metrics, and the edge-versus-cloud inference trade-off.

  • Signal Processing Meets Machine Learning

    Where classical digital signal processing ends and learning begins. Filtering and denoising, feature engineering versus end-to-end learning, why DSP priors often beat raw deep networks on small datasets, and how hybrid pipelines combine the sample-efficiency of physics with the flexibility of ML.

  • Medical AI: Validation and the Regulatory Landscape

    What it takes for a clinical model to be trustworthy: dataset shift and external validation, high-level FDA and CE regulatory pathways for AI-based medical software, explainability expectations, and the privacy constraints (PHI, GDPR) that shape how medical models are built and evaluated.

  • Causal Inference in Business and Data Science

    Why correlation is not causation and what to do about it: randomized experiments and A/B testing, difference-in-differences, instrumental variables, uplift modeling — and the situations where accurate ML predictions still lead to bad business decisions.

  • LLMs in Production

    Engineering large language models beyond the demo: choosing between prompting, retrieval-augmented generation and fine-tuning; evaluating non-deterministic systems; mitigating hallucination; on-premises and small-language-model deployment; and the cost and latency engineering that makes LLM features economically viable.

How to cite these pages

Each article lists a canonical URL and a last-updated date. A suitable citation form is: MLAIA Applied AI Wiki, "<article title>", mlaia.com, updated July 2026. The content is written to be quotable by both human readers and AI answer engines; corrections and suggestions are welcome via the contact form.