George Cann

AI at Carnegie Mellon

• DARE

  Co-created Carnegie Mellon's open-source AI platform for education. 2,000+ students served in courses from history to statistics, with partner institutions in the US and abroad piloting it.

  • One governed gateway to OpenAI, Anthropic, Gemini, Llama, Mistral and other open-source models, with role-based access and SSO.
  • RAG grounded in course materials, a visual workflow builder for multi-step teaching tools, and a research mode.
  • Usage, cost and per-message energy tracking, with audit trails for transparency.
  • Flagship of CMU's Open Forum for AI: open source (AGPL 3.0) and self-hosted with Docker, so any institution can run it on its own terms.
  ↗

• Exit Tickets

  After-class feedback platform: several hundred student submissions a week across multiple courses.

  • Compares student feedback with each lecture's learning goals; custom reports flag action items and gaps in understanding for instructors.
  • Gives students real-time feedback on their responses as they write, which has greatly improved response quality.
  • Homework: students submit work and are assigned peer grading, which is compared against AI grading.
  ↗

• Document pipelines

  AI extraction that turns archives and scanned PDFs into research data.

  USCT pension files

  • Transcribed and analyzed thousands of Civil War pension files of United States Colored Troops (USCT) soldiers with Gemini.
  • Used AI extraction and Splink record linkage to reconstruct the lives and stories of the people in the files.
  • Building a public website where people can find the stories of their ancestors.

  Historical sports records

  • Scanned old tournament record books and used Gemini to extract every amateur and professional match result from 1895 to 2024.
  • Analyzed how individual players rose through the ranks and were selected for bigger tournaments.
  • Built a web interface for browsing all 228 tournaments, 4,231 players and 177 venues, with full brackets of 64 and sometimes 128 players.
  • Transcribed a historian's interviews with the players and merged them into the dataset.

Data at Carnegie Mellon

• Leadership dashboards

  Teaching and business intelligence dashboards for college leadership, built in Tableau Server and later migrated to Power BI, fed by custom datasets and Python-automated data feeds.

• Finance reports and dashboards

  Built a full suite of finance reports and dashboards that help the finance team sort through and use large volumes of Oracle Financials data.

Data & analytics skills

• Visualization & BITableau (Desktop, Prep, Server), Power BI, Google Data Studio
• Data & warehousingSQL, Snowflake, Oracle Financials
• Programming & automationPython, automated data feeds
• AnalysisDataset design, data cleansing & integration, trend analysis, dashboard and report design

CodingPA LLC codingpa.com ↗

AI and data solutions built to run in production: custom AI applications, data infrastructure and software for researchers, businesses and organizations.

• fieldwork.cloud

  AI platform that turns documents, recordings and field devices into clean, structured data. Private beta.

  • Upload PDFs, notes and audio; AI extracts structured data and answers questions across all your files.
  • Connects Raspberry Pi and ESP32 boards with cameras and sensors (motion, weather, soil moisture), including solar-powered outdoor units with offline buffering.
  • AI watches camera feeds for the moments that matter, and simple automations turn events into actions, like motion → photo.
  • Custom dashboards chart sensor readings and file analysis side by side, or describe what you want and the assistant builds it.
  ↗

• AI trading simulator

  An AI research platform for public companies with a live AI paper-trading league. In production; open-source release coming soon.

  • Pulls SEC EDGAR filings, market data, Reddit sentiment and Wikipedia for each tracked company, then screens it on margins, free cash flow, growth and runway.
  • Citation-grounded chat assistant that reads a company's filings and your saved research.
  • Claude, GPT and Gemini agents compete with paper trades against live market prices, testing the screen against real decisions. No real orders or money.
  • Django, Celery and React, with screening and league agents running on a schedule around the clock.
  ↗

• Django + Next.js templates

  Open-source starters for launching secure full-stack apps.

  • Django backend: Google and email sign-in, WebSockets, Celery background tasks, invite-code signup and session revocation; Docker-ready.
  • Next.js frontend: session auth with the Django backend, CSRF-aware requests and a reconnecting WebSocket helper.

  Django template ↗ Next.js template ↗

Experience

• Senior Data Analyst & AI Developer · CMU2022 – now
• Principal Consultant · CodingPA LLC2025 – now
• Data Analyst · CMU2021 – 2022
• Senior Systems & Support Administrator · CMU2017 – 2021
• Support & Technology Consultant · CMU2011 – 2017
• Support Consultant · Solar Power Industries, Rostraver, PA2010 – 2011
• B.A., Environmental Studies · University of Pittsburgh2010
• Valedictorian · Union Area High School2005