AI sales management platform with LLM and RAG (Retrieval Augmented Generation)
- 6 months
- 2 persons
The project
B2B SaaS platform integrating generative AI to automate and optimize the sales process, from lead management to the generation of personalized business proposals.
Startup AI is an innovative startup specializing in the automation of sales processes. Their vision is to transform how sales teams interact with their clients by using AI to generate personalized and relevant business proposals.
Objective: Create an intelligent platform that automates the sales process
Challenges:
- Multimodal integration (voice, text, documents)
- Advanced personalization through AI
- Real-time processing of customer interactions
Key Features:
- Automatic call transcription
- AI-generated business proposals
- Vector memory system for customer context
- Automated personalized emails
- Multi-service integration (Aircall, AssemblyAI, OpenAI, Sigilium)
Technical proposal
Microservices Architecture in Go, utilizing:
- Event Sourcing for state management
- RAG (Retrieval Augmented Generation) for contextualization
- LLM Integration (GPT-4, Gemini, Mistral, Claude) for content generation
- Vector Embeddings for semantic search
- REST API for external services
- Webhook System for real-time integrations
Technical stack
- Go
- PostgreSQL
- OpenAI API
- Vector Embeddings
- AssemblyAI
- Aircall
- Sigilium
- Stripe
- Clerk
- Cloud-native
- Event Sourcing
- Event Driven
- RESTful APIs
Intervention areas
- UX/UI Design
- UI Development
- Architecture
- Backend Development
- Frontend Development
- Delivery Manager
- Proxy Product Owner
- Product Owner
- DevOps
- Operability
Results
- Automated call transcriptions
- 100%
- Reduced proposal writing time
- ~70%
- Calls handled per customer/day
- 450+
- Service availability
- 99.9%
- Integration with major external services
- 5
Our exFabrica touch
Event-sourcing architecture for complete traceability
Sophisticated prompt system for contextual content generation
Innovative multimodal integration (voice, text, documents)
Custom RAG approach for client memory
Dynamic guidelines for tone and format customization
Modular architecture enabling easy addition of new AI features