A full-stack analytics platform that brings together earned value management (ANSI/EIA-748), critical path scheduling, Monte Carlo risk simulation, DCMA 14-Point schedule quality auditing, and AI-driven insights — purpose-built for project controls professionals managing complex, high-value programs in construction, defense, and infrastructure.
EarnedAI is a comprehensive project performance intelligence platform that enables project managers, program managers, and portfolio stakeholders to monitor, analyze, and forecast project health across cost, schedule, and risk dimensions.
ANSI/EIA-748 Compliant
Full EVM calculations including PV, EV, AC, variances (CV, SV), performance indices (CPI, SPI), and forecasts (EAC, ETC, VAC). Supports Earned Schedule analysis with time-based SPI(t) and SV(t) metrics.
Forward & Backward Pass CPM
Custom-built CPM engine performs forward and backward pass calculations with cycle detection, identifying critical and near-critical paths (float < 10 days) across complex task dependency networks.
Stochastic Risk Modeling
Runs thousands of probabilistic simulations using PERT and triangular distributions to generate cost and schedule confidence intervals (P10, P50, P90), criticality indices, and sensitivity analysis.
Schedule Quality Auditing
Implements all 14 DCMA schedule quality checks — from missing logic and hard constraints to BEI and critical path integrity — producing a composite quality score with remediation guidance.
Rules Engine + Claude AI Integration
Rule-based AI engine surfaces cost overruns, schedule risks, and trend anomalies with severity scoring and recommendations. Claude Sonnet generates executive-level narrative reports (500-800 words) synthesizing EVM data, risks, forecasts, and WBS issues into actionable prose.
XER File Import & Mapping
Custom XER parser handles Primavera P6's tab-delimited export format, mapping activities, relationships, resources, and calendars into the application's type-safe data model for immediate analysis.
/Portfolio KPIs, S-curves, cash flow, cost breakdown, risk score, and performance gauges in a single view.
/projectsAll projects as interactive cards with health indicators, key metrics, and risk status at a glance.
/scheduleGantt-style timeline with critical path highlighting, earned schedule metrics, and task dependency visualization.
/schedule-qualityDCMA 14-Point pass/fail dashboard with remediation guidance and composite quality scoring.
/analyticsVariance waterfall charts, WBS-level analysis, cost breakdown, and performance indices radar.
/forecastingAI-driven EAC/ETC projections, trend analysis, scenario comparison, and completion date ranges.
/monte-carloHistogram distributions, confidence intervals, criticality index, and sensitivity analysis.
/ai-insightsNarrative insights with severity levels, recommendations, impact scoring, and category filtering.
/reportsAI-generated narrative reports via Claude, CSV data exports (metrics, tasks, WBS, insights), and copy/download functionality.
/loginSecure authentication with NextAuth v5, JWT sessions, role-based access, and demo account for evaluation.
EarnedAI is a monolithic Next.js full-stack application with clean separation between the presentation layer, state management, API routes, computation engines, and data layer. All proprietary calculation logic executes server-side, with the client handling visualization and interaction.
Built the EVM, CPM, and Monte Carlo engines from scratch rather than using third-party libraries — demonstrating deep domain knowledge and ensuring full control over correctness, performance, and extensibility.
Heavy calculations (EVM, CPM, Monte Carlo simulations) run on the server via API routes, keeping proprietary logic secure while letting the client focus on responsive, interactive visualization with Recharts.
Used React Context with useMemo caching instead of Redux or Zustand — appropriate complexity for the application's scope while still preventing unnecessary re-computation of expensive metrics.
Every component, engine, API route, and utility is fully typed with 40+ interfaces — catching errors at compile time and making the codebase self-documenting for other developers.
Built a custom Primavera P6 XER parser to bridge the gap between industry-standard scheduling tools and modern web analytics — handling the proprietary tab-delimited format with proper type mapping.
Chose SQLite with Write-Ahead Logging over PostgreSQL/Supabase — providing ACID-compliant persistent storage with no external dependencies, transaction support, and auto-seeding of sample data on first run.
Implemented NextAuth v5 with JWT sessions and a custom token-bucket rate limiter (100 req/min per IP) — securing the application without requiring a separate auth service while protecting API endpoints from abuse.
Configured a three-stage Docker build (deps, builder, runner) on Alpine — compiling native SQLite bindings in the build stage and producing a minimal production image with persistent volume mounting for the database.
EarnedAI V2 — Designed & developed by Eric Mitchell
Use the sidebar navigation to explore the live application.