Skip to content

Teaching Ecosystem Vision

The CSED Lab isn't building one tool. We're building an integrated platform for CS pedagogy — a suite of tools that share data, authentication, and design philosophy. This page explains how the pieces fit together.

The Problem

Teaching a large software engineering capstone requires solving several interconnected problems at once:

  1. Grading — How do you evaluate 39+ open-ended projects with unique requirements?
  2. Team Formation — How do you form balanced teams from 114+ students?
  3. Team Health — How do you detect struggling teams before they fail?
  4. Peer Evaluation — How do you capture team contribution fairly?
  5. Attendance & Participation — How do you know who's showing up and engaging?
  6. Identity — How do students authenticate across all of these without friction?

Each problem has been tackled independently, but the real value comes from connecting them.

The Tools

flowchart TD
    subgraph Foundation
        UMACS[UMACS<br/>Unified Auth]
        Canvas[Canvas LMS]
    end

    subgraph Assessment
        Lab40[Lab 4.0<br/>AI Autograding]
        GradeFlow[GradeFlow<br/>Visual Pipeline Builder]
        TeamEval[Team Evaluation<br/>Peer Assessment]
    end

    subgraph Formation
        Survey[Capstone Survey<br/>Student Profiles]
        SkillNet[SkillNet<br/>Skill Extraction]
        OptiTeam[OptiTeam<br/>Team Formation]
    end

    subgraph Engagement
        T4[T4<br/>Time & Task Tracking]
        Tendy[Tendy<br/>Attendance]
        UniPoll[UniPoll<br/>Live Q&A]
    end

    subgraph Infrastructure
        Nebula[csed-nebula<br/>GPU Compute]
        Psephos[Psephos<br/>Scoring Engine]
        TA[TA Copilot<br/>AI Assistant]
    end

    UMACS --> Survey
    UMACS --> TeamEval
    UMACS --> T4
    UMACS --> OptiTeam

    Survey --> SkillNet
    SkillNet --> OptiTeam
    OptiTeam --> Canvas
    OptiTeam --> Lab40

    Lab40 --> GradeFlow
    GradeFlow --> Canvas

    TeamEval --> Psephos
    TeamEval --> Canvas

    T4 --> Lab40
    T4 --> TA
    Tendy --> TA

    Nebula --> Lab40
    Nebula --> GradeFlow

How They Connect

The Data Pipeline

The ecosystem's core insight: every tool generates data that feeds the others.

  • Capstone Survey collects student preferences, skills, and availability → feeds SkillNet for skill graph inference → feeds OptiTeam for team formation
  • OptiTeam produces team assignments → feeds Canvas roster → feeds Lab 4.0 autograder's repository resolution
  • Lab 4.0 runs the staged evaluation pipeline → produces evidence traces → feeds back into GradeFlow as the first plugin bundle
  • Team Evaluation collects peer scores → feeds Psephos for sigmoid scoring → feeds Canvas gradebook
  • T4 tracks GitHub activity → provides effort metrics to Lab 4.0 for balance scoring → surfaces team health through TA Copilot
  • Tendy/UniPoll captures attendance and participation → feeds TA Copilot for engagement dashboards

The Identity Layer

UMACS provides a single sign-on across all tools. A student authenticates once and can: - Complete the Capstone Survey - Access Team Evaluation peer reviews - View T4 time-tracking dashboards - Log into OptiTeam for team formation preferences

No more Canvas LTI dance for every tool. This reduces friction for both students and instructors.

The AI Layer

TA Copilot acts as the natural language interface to the entire ecosystem. TAs and instructors ask questions in plain English and get answers synthesized across tools:

  • "Show me teams with low peer evaluation scores and high procrastination" → queries Team Evaluation (peer scores) + T4 (procrastination index)
  • "Which teams struggled with architecture diagrams this semester?" → queries Lab 4.0 grading snapshots
  • "Who hasn't checked in this week?" → queries Tendy attendance records

The Evolution

flowchart LR
    subgraph Phase1[Phase 1: Proof of Concept]
        Lab40 --> TeamEval
        Survey --> OptiTeam
    end

    subgraph Phase2[Phase 2: Integration]
        UMACS --> Identity[Unified Identity]
        Lab40 --> GradeFlow
        T4 --> Monitoring[Team Monitoring]
    end

    subgraph Phase3[Phase 3: Platform]
        GradeFlow --> Ecosystem[Ecosystem Platform]
        Psephos --> Scoring[Centralized Scoring]
        TA --> Agent[AI Agent Interface]
    end

    Phase1 --> Phase2 --> Phase3

Phase 1: Proof of Concept (Completed)

Individual tools proven in CSE120: Lab 4.0 autograding, Team Evaluation peer assessment, Capstone Survey + OptiTeam team formation.

Phase 2: Integration (In Progress)

UMACS provides unified identity. GradeFlow generalizes autograding into a platform. T4 adds team monitoring. Tools start sharing data instead of operating in silos.

Phase 3: Platform (Vision)

GradeFlow becomes the visual pipeline builder for any course. Psephos centralizes scoring algorithms. TA Copilot becomes the AI agent interface to the entire ecosystem. csed-nebula provides GPU compute infrastructure for all tools.

Why This Matters

The CSED ecosystem proves that teaching tools are not independent products. When grading, team formation, peer evaluation, attendance, and analytics share data and identity, the whole becomes greater than the sum of its parts. Instructors get actionable insights no single tool can provide, and students experience a cohesive platform rather than a collection of disconnected URLs.