AI Lab

Intelligence, applied with care.

Interactive demonstrations appear only where a real, safe endpoint exists. Everything else is clearly marked Simulation or Concept — no hidden models, no private data, no API keys in the browser.

01 · Agents

AI agents

conceptInteractive simulation

Agents that plan, call tools, and report what they did. Real-world counterpart: AI Jack and Ask Jack (see Projects). This demo is a scripted replay.

  • Tool use
  • Planning
  • Observability
Simulation

A scripted replay — no model, tools, or real data are involved.

  1. Press play to replay a representative agent run.

02 · Language models

Language-model integrations

concept

Embedding language models into products — cloud models with automatic local fallback, plus fully self-hosted inference. See AI Jack and PrinCio Nexus.

  • Prompting
  • Structured output
  • Evaluation

No demo yet — this area is a concept under exploration.

03 · Dashboards

Intelligent dashboards

concept

Dashboards that explain themselves — surfacing what changed and why it matters.

  • Analytics
  • Narrative summaries

No demo yet — this area is a concept under exploration.

04 · Retrieval (RAG)

RAG systems

conceptInteractive simulation

Retrieval-augmented answers grounded in documents and schema knowledge (vector search behind AI Jack). This demo runs real keyword retrieval, locally, over a small sample corpus.

  • Chunking
  • Retrieval
  • Citations
Simulation

Runs entirely in your browser over 6 sample passages. Real ranking, but no language model.

  1. [1] Citationsrelevance 100%

    A grounded answer cites the passages it used, so a reader can verify each claim against the source. If no passage supports a claim, the system should say it does not know.

    matched: do, verify, answer, against

  2. [2] Freshnessrelevance 14%

    Indexes go stale. Re-index changed documents on a schedule or on update events, and record the version of each passage that an answer used.

    matched: answer

  3. [3] Evaluationrelevance 13%

    Retrieval quality is measured with a labelled set of questions: did the right passage appear in the top results? Track recall and answer faithfulness over time.

    matched: answer

05 · Computer vision

Computer vision

conceptInteractive simulation

Detection and analytics over images and video. Real-world counterpart: face-verified attendance. This demo draws sample detections on a synthetic scene — no camera or model involved.

  • Detection
  • Tracking
  • Edge inference
Simulation

Synthetic shapes with made-up confidence values — no camera, video, or model.

06 · Automation

Workflow automation

concept

Event-driven pipelines that move work between systems without hand-offs — mailbox scanners, WhatsApp dispatching, spreadsheet sync, scheduled alerts.

  • Queues
  • Webhooks
  • Idempotency

No demo yet — this area is a concept under exploration.

07 · Future

Future experiments

concept

The open frontier — ideas queued for exploration.

No demo yet — this area is a concept under exploration.