Big Data Expert · Risk Control
10+ years in big data & distributed systems, 5+ years in risk control. Focused on real-time graph tech and loss-rate governance — cut subsidy loss rate from 10% to under 1%.
Self-built graph storage & query on HBase / HugeGraph; k-hop subgraph P99 ≤ 100ms for real-time risk decisions.
Graph strategies & features for brush / fake-order / content-safety / order-switching risks; cut annual loss rate below 1%.
Flink + Kafka real-time graph feature pipelines; Spark offline full-graph mining across onboarding & coupon risk.
Druid / ClickHouse / Doris real-time BI engine; PB-scale clusters powering company-wide dashboards.
Independent AI apps face a triple squeeze: features swallowed by model upgrades, growth that never turns into profit, and user entry points seized by upstream labs. Key takeaways from LatePost’s investigation.
My very first blog post, documenting how this site went from zero to launch.
A review of building this personal site with WorkBuddy (Ardot design + Astro + Giscus + GitHub Pages), and a reusable methodology and Skill distilled from it.
The full flow from project init to GitHub Pages deployment, including the pitfalls along the way.