Decisions, artifacts, metrics moved.

Problem

Brands measured impressions while real buying signals sat unmonetized in social feeds. No one had closed the loop from Twitter purchase intent to attributable QSR revenue.

Decision

Co-founded and made the core bet: real-time intent detection with assisted response over generic social publishing. Built intent scoring, human-in-the-loop approval, redemption tracking, and trust calibration so high-confidence signals moved autonomously while low-confidence required human review.

Problem

B2B teams burn six figures annually on meetings that move information instead of decisions. Context fragments across Slack, Jira, and docs with no synthesis layer — 4–6 recoverable hours per employee per week in a 50-person company.

Decision

Chose behavioral inference (response latency, participation, decision ownership) over semantic interpretation. Built explainability first, model sophistication last, to keep the product on the right side of the surveillance line.

Problem

Mid-market field operators in construction and retail run 50–200 reps with labor at 60–70% of cost and staff on gut feel. Incumbents captured task history but offered no predictive intelligence.

Decision

Architected the core product from vision to GTM. Sequenced three use cases — staffing forecasting first, sales presence second, AI rep training third. Built on existing LLM APIs instead of fine-tuning, accepting lower accuracy for 6-month faster time-to-pilot.

Problem

Research shows 20–40% of men on dating apps are married and self-reported profiles go unchecked. Generic apps see 30–40% activation and only 1–3% paid conversion — distrust kills engagement.

Decision

As fractional CPO, defined the thesis: verified signal, not self-reported data. Rejected income as a matching criterion (App Store risk) and designed AI verification quality-first — most failures are capture problems, not fraud. Vision, architecture, and roadmap delivered in 30 days.

Problem

Women 65+ with chronic back pain are underserved by every VR therapeutic on the market. No platform combined active attentional distraction, closed-loop biofeedback, and AI personalization tuned for this demographic's constraints.

Decision

Led product vision and MVP over 6 months. Delivered the vision at day 30 — still the team's operating foundation. Defined three pillars: visual foraging (active distraction), biofeedback, and AI personalization with an abstention-first policy (low-confidence outputs suppressed, never user-facing).