Bruno Ramael

CTO and AI / Full Stack Developer, I support startups and SMEs in creating SaaS products, AI agents, and high-value business automation solutions.

Technical founder of Eat&Work, a community building platform that automates engagement, matching, and professional event organization using artificial intelligence. I designed and developed the entire technical architecture as well as a conversational bot capable of connecting members based on their profiles, goals, and interests.

Key Skills:

• Python, FastAPI, Node.js, Next.js, React

• AI Agents and Generative AI (OpenAI, Mistral)

• RAG (Retrieval-Augmented Generation)

• ChromaDB, SQLite, PostgreSQL

• Chatbot and AI assistant design

• Business process automation

• APIs and microservices

• Observability and monitoring (Prometheus, Grafana)

• Cloud deployment and containerization (Docker)

Achievements:

• Development of an intelligent matching engine for professional communities.

• Implementation of a RAG pipeline using ChromaDB to enrich bot responses with business knowledge.

• Integration of GPT and Mistral models to improve recommendation relevance.

• Comprehensive monitoring architecture with Prometheus and Grafana to track performance and interaction quality.

• Automation of onboarding, events, and community engagement on Slack, Discord, and Telegram.

I prioritize projects where AI brings tangible business value: automation, internal assistants, customer support, recommendation engines, conversational agents, and AI-augmented SaaS platforms.

AI Architecture for Waste Detection and Segmentation – Edge & Cloud Distributed System

Design and development from scratch of a complete waste detection and segmentation platform for a startup specializing in environmental analysis, integrating computer vision, generative AI, and microservices architecture.

Context & Objective

Implementation of an intelligent system capable of:

capturing images via a Raspberry Pi in the field,

transferring data to the cloud,

automatically analyzing images to detect and segment different types of waste,

exposing results via a web platform and APIs.

Architecture & AI Pipeline

• Design of an edge-to-cloud distributed architecture based on Docker microservices.

• Implementation of an image processing pipeline:

capture on Raspberry Pi,

secure upload to Azure Blob Storage,

triggering of AI processing on the backend.

• Implementation of Computer Vision models:

Detectron2 for complex object segmentation,

Meta's Segment Anything Model (SAM) for generic and assisted segmentation.

• Comparison and selection of the most relevant model based on use cases (accuracy, robustness, real-world performance).

AI & LLM Innovation

• Design of a hybrid method combining Computer Vision and LLM to improve the detection of complex or ambiguous waste.

• Use of language models to enrich the interpretation of segmentation results and refine waste classification.

Backend & API

• Development of REST APIs in Express.js and FastAPI

• Containerized microservices orchestration with Docker.

• Deployment of services on Azure Container Apps / Azure Kubernetes (depending on project configuration).

Frontend Nextjs

• Implementation of a monitoring and observability system with Grafana

Intelligent Animal Detection for a Farm

Assisted a small agricultural company in implementing an automated surveillance system to detect the presence and movement of donkeys on the farm in real-time.

Achievements

• Design and development of an embedded computer vision system on Raspberry Pi.

• Creation and annotation of a custom dataset with CVAT for donkey identification in various contexts (weather conditions, lighting, viewing angles).

• Training and optimization of a YOLO model for real-time detection with embedded hardware constraints.

• Deployment of the model on Raspberry Pi with local video processing to limit infrastructure costs.

• Implementation of an automatic alert system via Telegram upon detection of specific events.

• Development of a Next.js web supervision interface allowing:

visualization of detections;

alert history;

equipment tracking;

activity statistics consultation.

Technologies

Python, YOLO, Computer Vision, CVAT, Raspberry Pi, Telegram Bot API, Next.js, React, Docker.

Result

Implementation of an autonomous surveillance solution based on AI, capable of detecting animals in real-time and instantly alerting the farmer, reducing manual checks and improving daily herd monitoring.