Builder of data-driven tools that turn raw information into smarter decisions

I build practical applications that aggregate data from multiple sources, detect patterns, and surface actionable insights — whether it's sports analytics, personal finance, or local fishing conditions. Python-first, automation-heavy, and always backed by real data.

Full-stack AI betting platform with an 11-agent analysis pipeline, per-sport ML models, and multi-user pick tracking.

• Tech: Python, Flask, SQLite (WAL), scikit-learn, APScheduler, Cloudflare Tunnel
• Sports: NBA (16 stats) · NHL (18 stats) · NCAAB (18 stats) · MLB (28 stats — Opening Day 2026)
• Highlights:
  • 11-agent pipeline (line movement, injury intelligence, trends, ML learning, consensus, narrative, and more)
  • Sport-siloed ML models — each sport trains independently, no cross-contamination
  • Dual-signal totals model (raw scoring + defense-adjusted) for Over/Under direction
  • Independent bet direction — spread, ML, and totals each compute separately
  • Shadow picks system — learns from all games, not just logged ones
  • Multi-user leaderboard with 30-day rolling window and H2H matchups
  • Auto-resolve scheduler (9:30pm + 3am EST), auto-scan (10am/2pm/6pm/10pm EST)
  • Nimrod — automated bet slip generator + Twitter/X posting agent
  • Real-time injury + ATS data from Covers.com
  • 6-key API rotation with auto-failover
  • Live at sportsbettingaianalyzer.com
• Status: Private repo (proprietary model)

6-year deep dive (2019–2025) into why teams win at every round of the NCAA Tournament.

• Coverage: All 192 R64 games, 96 R32 games, Sweet 16, Elite 8, Final Four, and Championship
• Key findings:
  • The "Efficiency Staircase" — each round has a distinct AdjEM floor teams must clear
  • R32→Sweet 16 is the steepest cliff in the bracket (+14 → +22+ AdjEM)
  • Champion gate: KenPom top 6 + AdjO top 25 + AdjD top 25 — unbroken in the KenPom era
  • "Extreme Teams" (elite one dimension, weak the other) = 0 championships in 22 years
  • Three Cinderella archetypes with distinct depth ceilings
• Live site: seang1121.github.io/ncaab-MarchMadness-Trend-analysis
• View Repository →

Portfolio strategy comparison tool across 30+ Fidelity mutual funds and ETFs.

• Tech: Python, GitHub Pages, GitHub Actions (auto-generates weekly reports)
• Strategies: 6 tiers from Conservative (5-7%) through Moonshot (60-150%+)
• Features: 1-year projections, risk profiles, rebalancing guides, DCA implementation plans, tax efficiency guidance
• Live site: seang1121.github.io/Fidelity-Fund-Analyzer
• View Repository →

CD ladder simulator comparing rates, strategies, and institutions to maximize savings yield.

• Tech: Python, GitHub Pages
• Institutions: Pentagon Federal, Navy Federal, Connexus, Vanguard, Ally, Marcus
• Strategies: 3-Rung Short, 5-Rung Classic, 5-Rung Staggered, Barbell
• Features: Maturity schedules, projected interest earned, liquidity timelines
• Live site: seang1121.github.io/CD-Ladder-Analyzer
• View Repository →

Self-contained NMLS SAFE MLO exam study course with structured learning across 4 modes.

• Modes: Topic deep dives, flashcards, quizzes, full practice exams
• Coverage: All 5 exam categories — Loan Origination (27%), Federal Laws (24%), Ethics (18%), General Knowledge (20%), State Content (11%)
• Features: 7-week study plan, 16 sessions, critical threshold reference tables, law-to-regulation mapping
• View Repository →

Daily mortgage rate monitor with historical analysis and trend detection.

• Tech: Python, Bankrate API + Chase scraper fallback, JSON storage
• Features: Day-to-day comparison, 7-day trends with directional indicators, configurable alert thresholds
• Use case: Rate shopping, refinance timing, market analysis
• View Repository →

6-source fishing conditions aggregator for Mayport, Jacksonville, FL.

• Tech: Python, BeautifulSoup, NOAA API
• Sources: ProAngler, FishingBooker, Tides4Fishing, NOAA Buoy MYPF1, TideTime
• Features: Live buoy data, tide predictions, solunar forecasts, water temp, wind conditions
• View Repository →

Multi-source aggregation — one source is a data point, six sources is intelligence

Pattern recognition — trends matter more than snapshots; why things happen matters more than what happened

Practical tools — everything I build solves a real problem I actually have

Clean architecture — sport-siloed learning, singleton DB patterns, no cross-contamination, automated scheduling

• Adding MLB to the betting analyzer (28-stat pitcher-first model, target: Opening Day 2026)
• Expanding NCAAB research with 2026 tournament data as it happens
• Refining multi-user weighted ML as pick history grows (75-100+ picks per sport)

• GitHub: @seang1121

"Where data meets decision-making"