See it work. Right now.
Four production demos that run against the real Bhala API. Real model, real sub-second latency, signed receipts on every call.
Counterfactual fairness audit
Paste any decision document. Algebraic de-racing + EEOC 4/5ths-rule disparate-impact ratio across six regulated domains. Intersectional analysis on the SCI manifold. Signed receipt.
Six domains · sub-second · signed
Open demoBias audit (28 protected axes)
Score any text on 28 BBQ + StereoSet + CrowS-Pairs + WinoBias axes. Per-axis drift, global Bias Score Index (BSI), top contributing axis. Signed receipt.
28 axes · v7 ChainMamba backbone
Open demoHate-speech classifier
9 per-category linear probes (women, LGBTQ+, jews, muslims, black, asian, latino, disabled, migrants). HateCheck AUROC 0.9339 — distinguishes 'I hate X' from 'Saying I hate X is bigoted'.
9 categories · 0.9339 AUROC adversarial
Open demoHallucination / grounding check
Given a model output and the source documents it was supposed to be grounded in, score how strongly the output is supported. Drop-in guardrail for RAG pipelines.
Q&A · summarization · dialog · signed
Open demoWant production API access?
The demo proxies above are rate-limited. For production-shape API keys (no rate limit, full audit-log access, SLAs), join the beta.