Jordan Beary
Data science leadership — analytics infrastructure, experimentation, marketplace ML
Bend, OR · linkedin.com/in/jordan-beary · github.com/JordanBeary
Download as PDF — this page and the PDF render from one source, so they cannot drift.
Detailed outcome figures are available in the full resume, via LinkedIn.
Summary
Data science leader with 8 years moving growth-stage companies from guessing to measuring. Founded and lead the Analytics & Data Science department at Trend, Inc. through a period of rapid company revenue growth.
Technical skills
Languages & tools: Python, SQL, R, Git, ClickHouse, Tableau, Holistics.io, Slack API
ML & AI: PyTorch, fastai, reinforcement learning & off-policy evaluation, NLP, LLM agent tooling, A/B testing & experiment design
Experience
Trend, Inc. (formerly Trend Capital) — Vancouver, WA
Analytics & Data Science Manager · May 2022 – Present
- Department creation. Proposed and won C-suite approval for a new analytics department; defined and hired the company’s first Data Analyst and Data Scientist roles.
- Revenue impact. Took three new business verticals through the full data maturity model — collection, reporting, prediction, automated optimization — with measured, material revenue impact.
- Experimentation. Lead A/B testing across verticals with rigorous design (power analysis, t-tests, chi-square, regression); own real-time bidding and auction mechanics.
- Leadership & communication. Set team priorities against company objectives; translate data science into recommendations actionable for both executives and engineers.
Performance Analyst · September 2019 – May 2022
- Analytics infrastructure. Built the analytics and BI platform for Client Services — descriptive through predictive reporting and real-time dashboards in SQL and BI tooling.
- Fraud detection. Designed fraud detection combining rules-based logic with unsupervised ML, measurably reducing fraudulent activity and chargebacks.
- Alerting. Created KPI alert systems (SQL, Python, Slack API) triggered by deviations from RNN time-series forecasts.
- Machine learning. Built lead sale-propensity models (fastai → PyTorch) used in production for re-marketing and fraud detection.
LegitScript — Portland, OR
Data Analyst · June 2018 – September 2019
- Marketing analytics. Automated conversion and customer-path reporting with Python and Google Analytics APIs, improving campaign targeting and conversion rates.
- Competitor intelligence. Built REST-API data pipelines with topic modeling, NLP, and sentiment analysis to map competitors’ market positioning.
- Risk modeling & visualization. Identified high-risk merchants with clustering in R; delivered Tableau geo-maps and 3D network visualizations of payment networks (Plotly, D3.js).
Data Research Intern · August 2017 – June 2018
- Data mining. Collected and analyzed data on thousands of pharmacies and manufacturers for targeted marketing; supported attribution modeling with statistical preprocessing.
Recent projects
- Outbound messaging optimization engine. Nightly scoring and suppression system for a high-volume outbound program (recency-based value thresholds, fraud-informed pre-send filtering); reversed a two-year decline in revenue per message and returned sold rate to a yearly high while cutting send volume substantially — profitable throughout.
- Fraud detection at scale. Identity-graph engine (union-find over hundreds of millions of records) with bot and emulator fingerprinting; attributed a major fraud surge, and now feeds alerting that keeps fraudulent leads out of the sold pool at scale while rescuing wrongly blocked funded customers.
- Closed-loop funding feedback pipeline. Daily ETL normalizing post-sale outcomes from partner lenders via privacy-preserving identity matching — per-partner funding rates measurable for the first time; feeds a funding-propensity model built on device-level signals.
- Reinforcement learning for lead routing. Auction simulator built from historical buyer behavior plus off-policy evaluation implemented from first principles; a transparent linear-program baseline validated a material revenue-per-lead lift on a large held-out sample, with the learned policy gated behind a live test.
- AI agent guardrails & knowledge base. Mediated, read-only data-access layer — query validation, domain-rule checks, capped results — enabling LLM agents to run analytics safely, paired with a version-controlled knowledge base read by humans and AI tooling alike.
Education
Portland State University — M.S. Systems Science · August 2017 – December 2019
Portland State University — Business Intelligence & Analytics Graduate Program · August 2017 – June 2018
University of Missouri – Columbia — Trulaske College of Business · August 2011 – May 2015