Engineering decisions
- A provider-agnostic AI layer keeps model choice configurable.
- Backend services separate business workflows from model calls.
- Human review sits between generated content and anything client-facing.
Back to selected work
Case study · 01
Compass
A production platform for client intelligence, content generation, workflow automation, and human review using multi-provider LLMs.
- Category
- AI marketing intelligence platform
- Availability
- Production system · Public case study
Problem + approach
Overview
Marketing teams need a dependable way to turn fragmented client context into structured intelligence, useful content, and repeatable production workflows.
Technology stack
- NestJS
- TypeScript
- PostgreSQL
- OpenAI
- Gemini
Conceptual overview
System architecture
- 01Client workspace
- 02NestJS API
- 03AI provider layer
- 04OpenAI / Gemini
- 05PostgreSQL
This project-specific overview is based on the published case-study architecture. Internal implementation, credentials, and deployment details remain undisclosed.
Decisions + constraints
Engineering notes
Technical challenges
- Maintaining consistent outputs across multiple model providers.
- Turning open-ended generation into reliable production workflows.
Results & impact
- Customer-facing AI workflows delivered at Apping Technology
- Client-specific metrics are confidential
Lessons learned
- Production LLM systems require strong orchestration and clear boundaries around generated output.
Screenshots + diagrams
Project media
01
Project media coming soon
Screenshots and diagrams will be added with the Compass case study.
Availability + roadmap
Documentation
Repository & API documentation
Future improvements
- Evaluation telemetry
- Expanded provider support