-
Use LLM Function Calling (Tools) in Modus Exploring how to integrate LLM "tools" (function calling) to build dynamic, AI-powered workflows.
-
AI-Driven Data Classification Using LLM (GPT) in Dgraph Leveraging LLMs to automate data classification and enrich knowledge graphs.
-
Add OpenAI, Mistral, or Hugging Face Embeddings to Your Knowledge Graph Practical guide to enhancing knowledge graphs with state-of-the-art embeddings for semantic search and classification.
-
Use OpenAI Embeddings in Dgraph for Auto-Classification How embeddings and Dgraph enable automated, scalable data classification.
- A Taxonomy of Knowledge Graphs for AI Use Cases Breaking down how knowledge graphs power AI applications, from search to recommendation systems.
Public case studies of projects I have implemented :
-
AgroPatterns: AI for Agriculture How we built AI-driven solutions for agricultural data analysis and decision-making.
-
How Pick Your Packer Built AI-Powered Search with Hypermode Case study on implementing AI-enhanced search for logistics and supply chain optimization.
-
OLAP Data Analysis with Dgraph: A Practical Guide Using Dgraph for analytical queries and real-time data exploration.
-
Rapid Application Development with GraphQL, Dgraph, and React.js Building full-stack apps faster with modern graph technologies.
-
Rapid App Dev: From Zero to Production Lessons learned from accelerating development cycles with Dgraph and GraphQL.
In 2018, I have used a mix of NLP techniques and a concept of ontology to create a "Conversational" user interface where people can just ask questions about their data in plain english. A pre-LLM project !
- Developed a natural language interface for querying complex datasets, combining NLP techniques and ontology-based graph databases.
- Key innovation: Used Dgraph for its flexible query language (DQL), enabling dynamic queries of interconnected business data.
- Use case: Demonstrated with airline flight operations data, allowing users to ask questions conversationally.
- Presented at a user conference in 2018.
📺 Demo: Conversational Query Engine 📺 Demo: Airline Data Use Case
Graph technology is an effective choice when you don't know in advance the type of queries and "joins" that you have to support. That's usually the case in all businesses where different data are interconnected.
At major events, we had a bike challenge and visitors could ride 30 seconds to do their best. Data is collected through sensors. We extended this experience : a spectator can wear Microsoft Hololens and look at the challenge and see the performance data ...
A quick explainer of the type of data that we have included in the demo:
By implementing this first innovation project I realized that using enterprise data in the context of Augmented Reality has some very valid uses cases.
The following is the introduction of the presentation I gave with a friend of mine at Global XR conference in 2020 on this subject.
The presentation also contained a "brief history of enterprise software architecture" which hopefully can help AR developers to better understand the enterprise software and data landscape (and it was fun to do !).
In 2020, our group has been asked to create an application that could help companies to get back to office safely given the COVID pandemic.
- Challenge: Build a COVID-safe office reentry solution in 2 weeks (prototype) and 3 months (production SaaS + PWA).
- Role: Led requirements, design, and architecture for a case management engine with contact tracing.
- Tech stack: AWS (S3, Route53, CloudFront, API Gateway, Lambda, DynamoDB), with VPCs and security groups for compliance.
- Result: A multi-tenant, scalable solution costing < $5/month to start, passing all security tests (pen testing, port scanning).
- Reused the serverless architecture to build a push notification system using SNS and DynamoDB streams.
- Multi-tenancy: Achieved via custom Lambda authorizers decoding JWT tokens for user organization context.