Validation
100%
Landing page and clickable validation experience complete
Portfolio Case Study · Delivery Complete
Roblox Decision Engine
AI Product & Growth Analyst for Roblox studios. The product turns retention, monetization and funnel signals into prioritized, evidence-backed actions and experiment decisions.
Roblox Open CloudProduct AnalyticsDecision IntelligenceFastAPIPostgreSQLAI AgentsExperimentationNext.js
Product
100%
Decision Engine, grounded AI Analyst and Experiment Copilot implemented
Engineering
100%
Roblox ingestion, decision path, model safeguards, security, reliability and production build are validated with CI and live provider smoke tests
Commercial
70%
Plans, validation pricing and 30-day guided pilot are defined; paid conversion remains to be validated with real studios
Product Thesis
Roblox tells studios what happened. The Decision Engine tells them what to do next.
The system validates metric changes deterministically, compares baselines and segments, generates a structured evidence bundle and only then lets the AI layer explain the finding. Supported findings can be turned into bounded experiments with explicit success thresholds and guardrails.
Architecture
Thin analytics layer. Strong decision layer.
01
Roblox Analytics
02
Query Adapter
03
Metric Normalization
04
PostgreSQL
05
Decision Engine
06
Evidence Bundle
07
AI Analyst
08
Experiment Copilot
Verified Evidence
The portfolio claim is backed by gates, not screenshots alone.
39
backend tests
Reliability gate passed with the full backend suite.
2
live providers
Roblox analytics and OpenAI model paths both passed live smoke validation.
0
frontend vulnerabilities
Dependency audit was cleared after the Next.js security upgrade.
15
launch routes
Landing, product demo, pilot, portfolio and public release metadata are included in the production route manifest.
Engineering principle
AI explains. Evidence decides.
Authoritative metrics and priority logic stay deterministic. The model receives structured evidence, cannot invent unsupported numbers or causal certainty, and falls back safely when validation fails.
Still unvalidated
Build complete ≠ market proven.
There are no claims of paying studios, recurring revenue, validated willingness to pay or proven KPI lift. That is why commercial validation remains at 70% even though engineering and portfolio delivery are complete.
Delivery