portfolio.sh
$ whoami
Victor Rios
$ cat role.txt
AI Engineer
$ ls ./skills
agents/ LLM/ computer-vision/ digital-arts/
$ cat vision.txt
Building intelligent systems at the intersection
of machine learning and creative technology
$

scroll to explore

$ cat about.txt

AI Engineer focused on building practical intelligent systems. I work across the stack — from fine-tuning large language models and training computer vision pipelines to deploying production-ready applications.

My background spans reinforcement learning, pose estimation, biometric systems, and generative AI. When I'm not engineering, I create digital art that explores the boundary between human creativity and machine perception.

Python PyTorch LLMs Agents GRPO / RL Computer Vision YOLO Pose Estimation UI design Digital Arts 3D / Maya
25+ Projects
3 Domains
AI Focus

$ ls ./projects

vision-rd — Pyronear wildfire smoke detection R&D 

                

                

                    
vision-rd

                    
Fork of pyronear/vision-rd — R&D on ML-based wildfire smoke detection, in collaboration with the Pyronear teams.

                    

                        YOLO
                        PyTorch
                        Pyronear
                    

                    

                        github ↗
                    

                

            


            
            

                

                    MoE Finetuning
                    

                

                

                    
MoE Finetuning

                    
Mixture of Experts model fine-tuning using GRPO reinforcement learning for optimized language model performance.

                    

                        Python
                        RL
                        GRPO
                    

                    

                        github ↗
                    

                

            


            
            

                

                    LLMcomparator
                    

                

                

                    
LLMcomparator

                    
Django-based platform for comparing LLM outputs side-by-side using OpenRouter API integration.

                    

                        Django
                        OpenRouter
                        LLM
                    

                    

                        github ↗
                    

                

            


            
            

                

                    YOLOv7 Weapon Detection
                    

                

                

                    
YOLOv7 Weapon Detection

                    
Real-time weapon detection system built on YOLOv7 architecture with custom-trained PyTorch models.

                    

                        YOLOv7
                        PyTorch
                        CV
                    

                    

                        github ↗
                    

                

            


            
            

                

                    AI Swing Analysis
                    

                

                

                    
AI Swing Analysis

                    
Computer vision pipeline for sports swing analysis using pose estimation and motion tracking.

                    

                        CV
                        Pose Estimation
                    

                    

                        github ↗
                    

                

            


            
            

                

                    Biometrics Face ID
                    

                

                

                    
Biometrics Face ID

                    
Facial identification system using biometric analysis for secure authentication and recognition.

                    

                        Python
                        Biometrics
                    

                    

                        github ↗
                    

                

            


            
            

                

                    LivePose Portal
                    

                

                

                    
LivePose Portal

                    
Real-time pose detection turned into interactive art — bridging computer vision and creative expression.

                    

                        Pose Detection
                        Art
                    

                    

                        github ↗
                    

                

            


        

    










    

        
$ ls ./gallery


        


            
            

                

                    PoseCafé — latte art generated by hand movement at the SAT Café
                

                

                    SAT · Metalab · LivePose
                    
PoseCafé

                    
Interactive installation at the SAT Café — become a latte art creator with a simple hand gesture.

                    
Reactive to body movement and the café's ambient sound. Built on LivePose, the SAT's open-source pose detection tool — a plug-and-play visual display driven by the surrounding acoustic and visual environment.

                    sat.qc.ca/fr/evenements/posecafe ↗
                

            


            
            

                

                    Mégaceta — immersive interactive installation at the Aquarium tropical de Paris
                

                

                    CREO · Aquarium tropical de Paris · HAR
                    
Mégaceta

                    
Collaboration with museum installation studio CREO — integrating a Human Action Recognition (HAR) system into an immersive marine experience.

                    
Exhibited at the Aquarium tropical de Paris, this work taught me to move from a lab prototype to a deployed installation — on-site calibration, visitor flow, reliability, and the operational constraints of a cultural venue.

                    creo.ca/fr/megaceta/ ↗
                

            


            
            

                

                    Panique dans l'espace — Planète Barilla VII, 3D environment for the web series
                

                

                    Web series · Blender · AI rendering
                    
Panique dans l'espace

                    
Collaboration on 3D set design for the web series — building sci-fi environments in Blender with AI-assisted rendering.

                    
Planetary landscapes and cinematic backdrops crafted for production, blending traditional 3D workflow with generative rendering to deliver immersive visuals for the show.

                    youtube.com — watch episode ↗
                

            


            
            

                
                

                    École Camondo · Robotics · Design
                    
École Camondo

                    
Collaboration on robotics design projects with interior architecture and design students.

                    
Bridging engineering and design — exploring how robotic systems integrate into spatial and product design workflows at one of France's leading design schools.

                    ecolecamondo.fr ↗
                

            


        

    










    

        

            
$ cat contact.txt


            

                github
                linkedin
                email
            


            

            
©  Victor Rios

        

    






    
    Gallery image