Peter Parker

Tells an organisation which customers, cases or patients to spend a limited intervention budget on: only the ones whose outcome the intervention actually changes. Retention offers, outreach, fraud review, clinical follow-up: a large share of every such budget goes to people who would have behaved the same way regardless, and this finds them.

uplift modelling · S/T/X/DR/R meta-learners and causal forests · budget-constrained policy optimisation · sensitivity analysis

Answers "which AI model should we use, and did the new version get worse?" with statistical confidence, using about a tenth of the evaluation calls a full benchmark needs. For a team re-evaluating models every month, that is a day instead of a week and tens of dollars instead of thousands, and it exposes the benchmark questions that were never measuring anything.

item response theory (2PL/3PL) · computerised adaptive testing · benchmark reliability and bias

No prompt or model change reaches users unless it is proven not to have regressed: the discipline a test suite gives code, applied to AI, with an audit trail a regulator can read. Includes a public month-by-month record of how the major vendors' "frozen" models silently change, which is the evidence most teams wish they had when something breaks.

LLM evaluation platform · judge calibration · confidence intervals and power analysis · CI regression gating · longitudinal drift

Monthly drift record begins September 2026.

Answers questions across thousands of financial filings where every figure is traced to, and checked against, the number the company actually filed; a wrong number fails loudly instead of reaching a decision. Cuts analyst hours, and eliminates by construction the invented figure that stops finance and legal teams trusting AI with documents.

retrieval over SEC filings · table-aware chunking · hybrid retrieval with reranking · XBRL fact-level grounding and automatic verification

Proves, with measurements, that a small model an organisation owns and hosts matches a frontier API on a high-volume task at a small fraction of the per-call cost, and shows exactly at what volume the switch pays for itself. For a team spending five figures a month on API calls, this is the project that finds most of it.

QLoRA fine-tuning · AWQ / GPTQ / GGUF quantisation · vLLM serving · quality, latency and cost frontier

Turns an access-to-information request from days of staff reading into hours of review: every proposed redaction carries the exemption category it rests on, a reviewer approves page by page with a full audit log, and the tool publishes how often it withholds what must go and, the number blunt PII tools never show, how often it keeps what must stay.

context-dependent PII redaction · exemption-category decision layer (rules plus LLM) · document layout parsing and OCR · per-category redaction confusion matrix · reviewer workflow with audit log

Forecasts demand (emergency-department arrivals, call volume, orders) with uncertainty ranges that actually hold when conditions shift, and turns them into a staffing number at a chosen service level. Overstaffing costs money and understaffing costs patients or customers; this puts a defensible number on both, site by site, rolling up to the region.

hierarchical probabilistic forecasting · neural forecasters (N-HiTS, PatchTST) against statistical baselines · MinT reconciliation · adaptive conformal intervals · CRPS and pinball loss · decision layer

Scores every transaction in under 50 milliseconds, thousands per second, before the money moves, and keeps working as fraud patterns shift, without the training-versus-production mismatch that quietly breaks most deployed models. The platform around the model is what organisations are actually missing, and this is a working one you can watch run.

streaming feature store · point-in-time correctness · online/offline parity tests · shadow deployment · drift-triggered retraining · expected-loss review queue

When a production system fails, this captures the failure with full context, drafts the fix, has a second AI review it adversarially, and hands an engineer a ready-to-approve pull request. Triage in minutes instead of days, engineers back on new work, and nothing changes without a person saying yes.

agentic incident triage · failure clustering · adversarial second-model review · human-gated patch proposals · confidence routing

For an inbox that receives hundreds of thousands of enquiries a year, a draft reply for each one, grounded in the organisation's own documents, with every claim cited to the passage that supports it and a clear escalation when the documents cannot answer. A person reads, edits and sends; the system sends nothing.

grounded response drafting · hybrid retrieval over a policy corpus · LangGraph triage, retrieve, draft, verify, escalate · citation verification · faithfulness measured with a calibrated judge · human approval before anything is sent

One product, configured per department in a few lines, that reads the day's news, public forums and social feeds, removes the duplicates, ranks what matters to that audience, and delivers a short digest with every claim cited and every quote verified against its source. One system in place of dozens of people each doing this by hand, at a few dollars a month per department.

multi-source ingestion (RSS, GDELT, Reddit API, Bluesky) · clustering and deduplication · per-audience relevance ranking · cited summarisation with quote verification · LangGraph pipeline on GCP Cloud Run · precision and recall against a labelled week