The challenge
The AstroSure team had a validated concept — daily astrology guidance powered by modern AI — but their first prototype could not scale. Readings were slow, prompts were hard to tune without redeploying code, and there was no reliable way to manage users, subscriptions, or content moderation.
They needed a founder-friendly engineering partner who could:
- Design a Django backend that could serve both web and mobile clients from one API layer
- Integrate LLM pipelines for personalized horoscope and chat responses without runaway token costs
- Ship an admin system the non-technical team could use to adjust copy, monitor usage, and review AI output
- Deploy on AWS with observability, background jobs, and room to grow past the first 10k users
Our approach
SparkScribe architected AstroSure as a modular Django platform with a clear split between user-facing APIs, AI orchestration, and admin operations.
Backend & API layer
We built REST APIs for user profiles, birth charts, daily readings, panchang data, and the Ask Agastya chat experience. Redis caching reduced repeat LLM calls for common daily content, while Celery workers handled long-running generation tasks asynchronously.
LLM pipeline
Prompt templates, safety filters, and structured output parsing were managed through Django admin — so the product team could iterate on tone and reading format without developer involvement for every copy change.
Admin & operations
A custom Jazzmin-backed admin gave the client visibility into user activity, reading logs, failed jobs, and feature flags. We added role-based access for support staff vs. product owners.
Delivery approach
We shipped in two-week iterations: core auth and chart engine first, then daily reading flows, then the Agastya chat interface and mobile-optimized endpoints. Weekly demos kept stakeholders aligned across US and India time zones.
Results & impact
AstroSure launched on schedule with a backend the team could operate independently. Early user feedback highlighted fast daily readings and a chat experience that felt personal rather than generic.
- Performance: Cached daily content and async workers cut average reading generation time by roughly three compared to the original prototype.
- Reliability: Production APIs maintained 99.2% uptime across the first major marketing push.
- Engagement: Ask Agastya became the most-used feature within the first month, driving repeat daily sessions.
- Maintainability: The client team updates prompts, copy, and feature toggles through admin — no redeploy required for most content changes.
SparkScribe continues to support AstroSure with monitoring, performance tuning, and new feature development as the product scales.
Additional overview
AstroSure.ai is a consumer-facing astrology platform that combines traditional chart calculations with LLM-generated narratives. Users receive personalized daily guidance, panchang calendars, compatibility insights, and a conversational AI guide named Agastya.
SparkScribe owned the full backend: Django services, database design, third-party astrology data integrations, OpenAI orchestration, subscription hooks, and the admin tooling the client uses to run the business day to day.
Implementation details
Technical highlights
- Django 5 REST API with token-based auth for web and mobile clients
- PostgreSQL for user data and reading history; Redis for cache and Celery broker
- Structured LLM outputs validated before persistence — reduces bad responses reaching users
- S3-backed media storage for user uploads and generated report assets
- CloudWatch logging and error alerting integrated from week one
Why Django
The client prioritized speed to market and long-term maintainability. Django's admin, ORM, and ecosystem let us deliver a production-grade backend quickly while keeping the door open for complex business rules around readings, subscriptions, and regional content.
