01. Abstract & The Performance Bottleneck
In airline operations, computing the cost of an isolated flight delay is trivial; modeling the multi-day network cascade it triggers is an expensive computational bottleneck. SkyRipple is a full-stack discrete-event delay propagation engine and agentic recovery simulator.
Operating on 582,304 real BTS flight records, 190,711 crew pairings, and 2.41 million synthetic passenger itineraries, the core engine proves that a localized disruption (e.g., a 2-hour capacity drop at ORD) carrying a direct cost of $181,376 triggers a network-wide cascade of $942,080 - a 5.19x propagation multiplier.
The Optimization
By isolating state mutation from graph traversal and aggressively monkey-patching functools.lru_cache onto the passenger loaders in Python, itinerary generation dropped from 110 seconds to 8 seconds. The live FastAPI endpoint achieved a steady-state resolution time of 16-20 seconds for a full national cascade.
167MB SQLite DB
Immutable Dec 2025 Schedule
16-20s Resolution
Full national cascade
5.19x Multiplier
Delay propagation cost
02. Agentic Arbitration & The LLM Boundary
To autonomously recover from cascading disruptions, SkyRipple deploys a 5-role Operations Control Center (Aircraft, Crew, Passenger, Gate, Duty Manager). Traditional airline recovery systems rely on greedy rule engines; SkyRipple replaces this with arbitrated financial simulation.
Gemini Proposes
gemini-flash-lite operates purely as a routing layer in /api/qa_router.py. It maps complex inputs ('delay United 1234 by 90 mins and ground N24993 for 4 hours') to a strictly validated Pydantic schema.
Code Disposes
The LLM never returns a final calculation. The structured intent is passed to deterministic code which runs pure functions against the pre-computed engine state.
Duty Manager Agent
Intercepts proposals, runs isolated financial simulations on each branch, blocks destructive swaps, and executes the mathematically optimal reassignment. Outperformed greedy logic by exactly $172,854 on a single disruption.
OCC Multi-Agent Arbitration
03. The Multi-Day Midnight Seam & Guardrails
Airline schedules do not reset at midnight; delays carry over in the form of displaced aircraft, illegal crews, and missed connections. Phase 8 of SkyRipple implements a continuous operational month engine.
Continuous Physics
The engine models tight-gap overflows. For example, tail N464UA carried a 335-minute unrecovered delay through a 395-minute overnight maintenance buffer.
Spillover-Aware Costing
Standard bounded calculations arbitrarily cut off reactionary costs at midnight. Spillover-aware calculation captured the true value by accurately crediting downstream delay reductions that calendar-based boundaries discard.
Honesty Guardrails
If the engine finds no net-beneficial actions, the system suppresses all celebratory UI states and plainly declares doing nothing was optimal. Hallucinated 'phantom savings' are structurally impossible.
04. Deployment Architecture: SkyRipple Lite
Serverless Precomputation
Instead of running the Python engine live on the web, massive multi-day scenarios are precomputed locally and exported to a zero-latency Next.js static client.
Browser-Side Analytics
Multi-gigabyte JSON outputs were partitioned by domain. The frontend fetches these dynamically in parallel. All dynamic scaling and exports execute natively in the client browser in 0ms, delivering full analytical depth with zero infrastructure overhead.