About
Revolutionizing how US service providers generate personalized business offerings through an automated platform using LLM and a vector database to manage products and clients.
Challenge, approach, and impact
Automating Business Product Recommendation System
Developing a cutting-edge system using an LLM integrated with a vector database to manage 100,000+ products and 50,000+ clients. The system delivers precise, scalable recommendations tailored to customer needs, calculates forecasts, and potential revenue.
Challenges & Solutions
1. Challenge: Deliver user-product matches. Solution: Well-trained LLM, testing different prompts, and implementing the most suitable one. 3. Challenge: Maintain speed in creating matches. Solution: Implemented a queue and parallelization for faster response times. 4. Challenge: Growth Dashboard and Simplex Admin for customer, products, and rules management.
Client's End-to-End Solutions Demonstration
Demonstrating the system's capability to deliver customized solutions for specific customers. Providing personalized product recommendations with detailed explanations to enhance transparency and trust in the decision-making process.
Solution 1
Dataforest developed an Admin account functionality that includes basic customer search, product recommendations with reasoning, recommendation score, and product image display, and the ability to modify the reasoning text.
Challenge 2
Functional requirements for Simplex team admins, including setting up new organizations, CRUD operations for product catalog and details, defining business rules, and viewing organization prompt logs.
Solution 2
To be filled with specific solution details.
Deployment and System Architecture
Good separation of DB, server, and UI. Deployment to AWS based on Kubernetes. CI/CD infrastructure and QA automation framework. Multi-tenant system with basic login/registration. Import/Export functionality for Products and Customers.
Recommendation Logic
Building a recommendation engine based on an LLM agents framework, utilizing org prompt templates, customer details, product descriptions, business rules, and generating recommended products with reasons and classifications.
Generative Recommendation System
Developing a system to provide recommendation scores, product images, thumbs up/down options for each recommendation, and the ability to provide feedback and save new business rules.
Results
Achieved cost savings, created a unique startup idea applicable to various industries, scalable to support thousands of users.
How we built
Testimonials
Anonymous
Dataforest
“Built an LLM-powered recommendation platform that scales personalized business offers across complex domains. The solution processes over 100,000 products and serves tailored recommendations to 50,000+ customers in under one minute. By combining RAG, vector search, and advanced prompt engineering, the platform delivers highly accurate recommendations, revenue forecasts, and actionable insights that significantly improve sales efficiency.“
Team structure
Client team
N/A
Investment Manager
Project stakeholder
The client stakeholders were working closely with the team at Dataforest
Agency team
1 x Tech Lead
Production
1 x Product Manager
Governance
1 x Business Analytics
Production
3 x Machine Learning Engineer
Production
