Infrastructure · 2012–2017
AccelaStudy Cloud
Four generations of cloud. One mission: sync what you learn.
Read the Full StoryEvolution
Four Generations
From zero lines of server code to a fully-owned AWS infrastructure — each generation was a response to a platform dying, a scale threshold breaking, or an ambition outgrowing the architecture beneath it.
Renkara Cloud v1.0.0 — August 2012
Built on Parse (Facebook's BaaS). MongoDB on mLab. Facebook and Twitter authentication. Cross-device progress sync between iPhone and iPad. User-generated vocabulary stored in the cloud. Over 1 million cloud accounts by end of 2012.
Parse + Firebase Hybrid — October 2015
Parse retained for identity and storage. Real-time sync moved to Firebase for millisecond-level propagation between simultaneous clients. A transitional architecture — two third-party dependencies for one feature.
Custom Node.js on AWS — May 2016
Parse shutdown forced full migration. Custom Node.js API at cloud.renkara.com. MongoDB replica set on mLab. Scheduled user import from Parse (every 5 minutes). Firebase eliminated February 2017 with custom real-time sync engine. 90,000 accounts by February 2017.
AVIAN Engine Subsystem — 2025
Cloud infrastructure absorbed into the AVIAN adaptive learning engine. Python 3.12+ / FastAPI / PostgreSQL / Redis on AWS. The sync layer is now one subsystem among 33, powering knowledge graph traversal, proficiency modeling, and adaptive assessment.
Capabilities
What AccelaStudy Cloud Did
The technical capabilities that powered hundreds of thousands of learning sessions across devices, networks, and continents.
Cross-Device Progress Sync
Study on your iPhone during the commute, pick up on your iPad at home. Every flashcard flip, mastery score, and study session synced automatically between all devices in real time.
Complete Offline Support
Download content at home on WiFi, study anywhere without an internet connection. Multi-master replication ensured all offline changes merged cleanly when connectivity returned — no data loss, no conflicts.
User-Generated Content
Users could create their own vocabulary words, study sets, and flashcard decks and store them in the cloud. The most-requested feature in AccelaStudy history — it drove a significant increase in engagement and downloads.
Multi-Master Replication
Not client-server sync — true multi-master replication. Any device could write at any time, online or offline, and the system would converge to a consistent state. No "please wait for sync" dialogs.
Single-Digit Millisecond Latency
Global latency under 10 milliseconds for API responses. Study sessions felt instant because they were. The architecture was optimized for the reality that learners don't wait — they swipe.
Social Authentication
One-tap signup via Facebook or Twitter (Generation 1) or email with verification (Generation 3). Minimized friction between "download the app" and "start studying." Account creation took under 10 seconds.
Aggregate Learning Analytics
Cloud-stored study data enabled aggregate analysis across all users. The system identified which vocabulary words users commonly struggled with and which were simpler — feeding back into the AccelaStudy algorithm to accelerate learning for everyone.
Premium Subscriptions
Free tier for basic sync, paid accounts at $12/year ($1/month) for premium features. By end of 2014: 100,000 paid subscribers generating over $1.2 million in annual recurring revenue.
COPPA, GDPR, and PCI Compliance
Children's Online Privacy Protection Act, General Data Protection Regulation, and Payment Card Industry compliance built in from Generation 3. Education software used in schools needs to meet these standards — AccelaStudy Cloud did.
100% Owned Infrastructure
After the Parse shutdown lesson, Generation 3 used only MIT-licensed dependencies. No proprietary third-party lock-in. The entire stack — API, sync engine, data layer — was fully controlled by Renkara.
Technical Specifications
Under the Hood
Generation 1 (2012)
- Parse SDK (iOS)
- MongoDB on mLab
- Facebook and Twitter OAuth
- Polling-based sync
- Zero server code
Generation 2 (2015)
- Parse for identity/storage
- Firebase Realtime Database
- Sub-second sync propagation
- Hybrid two-service architecture
Generation 3 (2016–2017)
- Custom Node.js API (7,000 LOC)
- MongoDB replica set on mLab
- AWS infrastructure
- Custom real-time sync engine
- Token-based auth with hourly cleanup
- Staging + production environments
- Parse user import (5-min intervals)
Generation 4 (2025+)
- Python 3.12+ / FastAPI
- PostgreSQL + Redis on AWS
- gRPC engine (119K LOC, 33 subsystems)
- RS256 JWT authentication
- Stripe subscription billing
- Knowledge graph + proficiency modeling
Scale