Team alignment is a science, not a gut call.
GitSyntropy is a multi-agent AI system that combines GitHub behavioral telemetry with psychometric profiling to predict how well software developers will collaborate — before the first sprint begins.
The Problem
Engineering team failures rarely come from individual skill gaps. They come from misaligned working styles, communication friction, and chronotype conflicts that are invisible to recruiters, managers, and even the engineers themselves until weeks into a project.
Our Approach
We built a four-stage LangGraph pipeline: a GitHub Analyst agent extracts commit timing, PR velocity, and collaboration patterns; a Psychometric Profiler builds an 8-dimension behavioral model from an adaptive assessment; a Compatibility Engine scores pairwise and team-level alignment across those dimensions; and finally Claude Sonnet synthesizes everything into a structured, actionable report with concrete hiring recommendations and workflow adjustments.
How It Works
GitHub Analyst Agent
Ingests commit timestamps, PR open/close rhythms, issue labels, and collaboration graph edges. Derives chronotype (owl / lark / balanced), activity rhythm score, and collaboration index.
Psychometric Profiler
An adaptive 8-dimension assessment (Innovation Drive, Leadership Orientation, Risk Tolerance, Stress Response, and more) builds a behavioral fingerprint for each team member.
Compatibility Engine
Scores all N×(N-1)/2 member pairs across 8 weighted dimensions. Aggregates to team-level weak, strong, and at-risk signals. Falls back gracefully when data is incomplete.
Claude Synthesis
Claude Sonnet streams a narrative report covering team alignment summary, friction risks, recommended meeting windows, hiring gap analysis, and async workflow adjustments.
Built With
LangGraph
Multi-agent orchestration graph
Claude Sonnet 4.6
Narrative synthesis + structured output
GitHub API
Behavioral telemetry extraction
FastAPI + Python 3.12
Async backend on Render
Astro 4 + React
Hybrid SSR/CSR frontend on Vercel
Supabase (PostgreSQL)
User profiles, assessments, scores