Ismael Goulani

L'IA générative et agentique a multiplié les slides et les preuves de concept (PoC), mais la mise en production reste le vrai point de bascule et c'est là que la plupart des projets calent. Fort de ce constat, Fynq conçoit, déploie et exploite vos agents IA de bout en bout, en managed service. Pensé pour les startups, scale-ups et PME, vous ne pilotez pas des agents : vous en récoltez le résultat, opéré et mesuré en continu. Nous pouvons nous engager sur un périmètre, un prix et un délai fixes, là où un cabinet facture au temps passé, parce que chaque mission s'adosse à notre plateforme, Fynq Agents Cloud : orchestration, supervision, journal d'audit, mesure d'impact, intégrations. Le socle qui rend le déploiement répétable, garanti et livrable en semaines, pas en trimestres. Trois formats productisés :  → Blueprint : diagnostic chiffré et feuille de route (5 jours)  → First Agents : 2 premiers agents en production en moins de deux semaines  → Factory Deployment : agents illimités sur une plateforme unique Chaque mise en production est livrée par nos forward deployed engineers seniors. Hébergement en Union européenne, avec déploiement dans votre propre environnement si vous l'exigez. Stack : Temporal, Pydantic AI, event sourcing, ClickHouse, architecture multi-fournisseurs.

As a data scientist at Axionable AI, My role was to design and deploy data products at the Edge or in the cloud. Fields: Deep Learning, Edge computing, computer vision, data analytics Tools : Tensorflow, Pytorch Aws Sagemaker, Docker, Python, Node.js, Pandas AWS cloud, Raspberry pie

GarageISEP is ISEP's very own innovation hub! In just one year it has grown from 4 founding members to 36 administrative members, impacting almost 1000+ students each year throughout hackathons, workshops, projects, and conferences. The structure is made of 5 Labs covering leading innovative technologies in AI, Robotics, Augmented & Virtual reality, Block-chain, Cyber-security, and Advance Algorithmic. The structure has established partnerships with major sector leaders such as IBM, Amazon, Capgemini, BearingPoint, Thalès, IEEE and more! During my role as Mentor and technical Advisor at IALAB I carried out the following tasks: ▪ Creation of educationnal content on machine learning (videos, cheat sheet...) ▪ Project management (autonomous car, football player price prediction...) ▪ Organization of workshops (scrapping, deep learning) ▪ Organisation of a hackathon with more than 100 people, sponsored by AWS ▪ Settings of technical goals for the projects ▪ Promotion of AI development within the school

Under the supervision of Dr David Rossiter, the topic of my research project was to use the QRNN(Quasi reccurent neural network) architecture to classify 8M Youtubes videos Technical Stack: Python, Tensorflow, GCP

When a vehicle is damaged, the underwriter of a policy can ask his insurer to obtain compensation for his injury and have his vehicle repaired. In particular, an expert may have to move and this generates a slowdown in the procedure as well as additional costs. During this internship, the goal was to automate this process of expertise through the use of Deep Learning algorithm. From a photo we must: ▪ Recognize parts of a vehicle ▪ Detect if part of the vehicle is damaged ▪ Propose a list of parts needed to repair the damage as well as an estimate of the possible cost of repairs ▪ Benchmark the proposal with other quotes received by AXA and approved by human operators. Within the Enterprise Content Management (ECM) Team , I was involve in all the main steps of the project. From the analysis of needs to the development of the tool itself, through the conception of the algorithms and the software. Among the tasks, we can mention : ▪ Benchmark and study of commercial and open source solutions of artificial intelligence ▪ Identification and organization of meetings of potential commercial partners for the realization of the project (Startups specialized in IA, IBM, AWS) ▪ Organization of meetings with Deep Learning specialists from the AXA research center in Lausanne ▪ One-off training of ECM team members on the different concepts of Deep Learning and Machine Learning ▪ Creation of scripts for the acquisition of image data on the net (Web Scraping) ▪ Pre-processing of the data ▪ Establishment of an AWS Cloud Work Environment for Training and Deployment of Deep Learning Models ▪ Development of Deep Learning architectures for vehicle damage estimation ▪ Creation of APIs to serve the different models of Deep Learning created ▪ Creation of a web interface to make demos with AXA IT decision-makers Technical stack : AWS, Python, Flask, Jupyter notebook, Scikit-Learn, Opencv, Pandas, TensorFlow, Keras, Selenium