Olivia building

Overview. In plain terms: Olivia tells an investor client which properties are likely to become a deal they can actually win — not just any home that sells. Olivia identifies which properties will become a dealable transaction for an investor client (wholesale, fix-flip) — narrower than "any sale." Spine = funnel decomposition: P(dealable) = P(transacts) × P(dealable | transacts). Stage A (transacts) leans on abundant deed labels (the generic transact signal, cf. Apollo); Stage B (dealable | transacts) trains on the client's closed deals as positives under positive-unlabeled learning, corrected with a true prior. A standing value-audit harness proves the as-of-T0 ranked list recalls real deals vs Alpha and the market base rate; a cadence layer sequences each client's monthly list. Gated upfront by a one-time data audit that freezes the label spec.