Banking and financial services
Real-time fraud scoring per transaction, AML sanctions screening against a 10,000-pattern blocklist with categorical inhibition on fuzzy variants, KYC pipeline running entirely client-side, RM cohort analysis on 10M customer records with behavioural filters, multi-rail payments tracker, on-device customer assistant (NBK Whisper), and in-browser file pre-flight scanner (SafeUpload) for iBusiness uploads. Hadron is the only stack where customer PII never leaves the user's device, automatically satisfying CITRA Tier 3/4, CBK CORF, and PDPL residency rules.
Observability and monitoring
Browser-side dashboards on millions of trace events. Filter, group, sort at panel-refresh speed. Multi-pattern log search with regex refusal.
Faceted search SaaS
Sub-keystroke refinement across half a million products. Continuous slider updates without debounce. The InstantSearch.js replacement.
In-browser analytical SQL
NYC taxi explorer on 1M synthetic rides. Cohort builder with live result-size feedback. A/B test analyser running entirely client-side.
Workspace search and productivity
Workspace search across 100K synthetic documents with no degradation as the workspace grows. Activity feeds with instant mention filtering.
In-browser ML inference
Live translation with a WebGPU vs WASM backend toggle. Sentiment classification stream. Semantic search RAG without a vector database.
ML experiment tracking
Filter / group / sort 5,000 simulated runs at any scale. Loss-curve overlays of fifty runs in a single frame.
Browser code editors
Find-in-files across a 20K-file repo, an in-browser secret & vulnerability scanner that replaces a Snyk subscription, and an AI completion re-ranker that scores candidates locally instead of paying for an LLM judge call.
Trading and financial dashboards
Streaming order book at 1,000 ticks per second alongside a live-aggregating portfolio P&L panel. 60fps holds while a naive baseline craters to 30.
On-device training and personalisation
An email-urgency classifier and a news recommendation tuner that both train fully in-browser. Every gradient step picks its own backend; the recommender re-ranks the catalogue with compute + sort engines after every thumbs up.