Romain CREUZENET

🔹 Web3 Wallet Design and development of a Web3 wallet aimed at a mainstream audience. I expanded the initial product vision to include an integrated messaging system and community hubs inspired by Discord (public and private). • Built a basic messaging system (SQL storage, prototype version) • Community hubs: creation, management, interface • Secure authentication via JWT signed by external wallets (Metamask or Phantom) or password • Functional prototype tested internally • Initial frontend development (POC), later handed off to a dedicated team 🔹 Crypto Swapper Design and development of an integrated crypto swapper. Originally based on Uniswap V3 and forks (e.g. Pancake), with an extensible architecture allowing computation of optimal swap paths (customizable number) across multiple DEXs. • Custom algorithm to compute top swap paths (no overlap), handling gas and hop constraints (Dijkstra and A* unsuitable) • Developed a Rust module for high-performance route computation • Price extrapolation using optimized sampling on a logarithmic scale to model Uniswap V3 liquidity curves • Real-time blockchain monitoring for pool price changes • Automatic detection of newly created pools • Fine-tuned RPC query optimization • Smart contracts in Solidity: multi-DEX swap with fee capture in a single transaction • Handling of atypical tokens and unstable price behaviors 🔹 Shared Tech Stack • Frontend: Angular (TypeScript), Metamask & Phantom integration (provider, account/chain switching), POC implementation • Backend: FastAPI (Python) + REST API, SQLAlchemy • Blockchain: web3.py · Solidity · Hardhat · deployed on Ethereum Reference: Bertrand Perrin — linkedin.com/in/bcperrin

Contributed to the maintenance and optimization of a large-scale backend system aggregating marketplace data flows for major companies such as Total, Veepee, and other high-volume partners. The platform handled millions of live orders, products, and merchant interactions, with strict requirements for uptime and scalability. 🔹 Platform Maintenance • Ensured production-level security, performance, and availability • Proactive resolution of incidents and anomalies • Continuous system updates to maintain large-scale stability 🔹 Data Management & Scalability • In-depth database analysis to identify performance optimizations • Data cleanup and archival strategies for cost and efficiency control • Large data volume handling to ensure responsiveness and cost-effectiveness 🔹 Tech Stack • Python / Django, SQL, high-volume big data environment

Design and development of a blockchain-based game platform — https://www.orilys-conquest.com Players and early investors could buy the game's cryptocurrency and NFTs, with resale directly available on the platform. An affiliate system was integrated to boost visibility. The UI was specifically designed to remain accessible for crypto newcomers, while offering a smooth and ergonomic blockchain experience. 🔹 Frontend • Angular (TypeScript) • Metamask integration: provider, account and chain switching • Game-oriented components focused on conversion and onboarding 🔹 Backend • Django (Python) + REST API, SQL database • Blockchain interaction via web3.py • Management of users, in-game resources, and rewards 🔹 Blockchain • Smart contracts in Solidity: ERC20, ERC721, marketplace, minters • Development with Hardhat & Remix IDE • Deployment on Mumbai (testnet), then Polygon (mainnet)

Design and development of a full training environment for a neural network aimed at detecting fake ID cards from image analysis. 🔹 AI Modeling & Training • Dynamic preprocessing, multi-dataset management • Optimization for constrained servers (multi-network setups, memory tuning) • Architectures: ResNet, EfficientNet • Training with PyTorch Lightning on Docker server 🔹 Web Demo • Web app allowing users to test the AI by uploading images • User authentication, confidence score display for each analysis 🔹 Fullstack Tech Stack • Frontend: Angular • Backend: Django (Python), PostgreSQL • Deployment: AWS EC2

Full design and development of a crypto-enabled e-commerce platform, accepting cryptocurrencies (ETH, USDT, etc.) with automatic conversion to euros for supplier payments. Included multi-chain support, floating exchange rates, and an optimized user experience. 🔹 UX/UI • Optimized checkout: group purchases, bundles, discounts • Smooth and simplified user journey for non-technical profiles 🔹 Frontend • React (TypeScript) • Metamask integration (provider, account/chain switching) • UI management of multiple blockchains and cryptocurrencies 🔹 Backend • Django (Python) + REST API • Robust database architecture • Blockchain integration using web3.py 🔹 Blockchain • Smart contracts in Solidity: ERC20 • Development with Hardhat & Remix IDE • Deployment on Mumbai (testnet), then Polygon

Design and development of an e-commerce platform specialized in the sale of resin-printed figurines. Users could submit their own models to expand the catalog, promoting a creative and community-driven approach. 🔹 UX/UI • Optimized checkout flow: fewer steps, bundles, special offers • Smooth, conversion-oriented navigation experience 🔹 Frontend • React (TypeScript) • Responsive layout with catalog and product detail views 🔹 Backend • Django (Python) + REST API • Full database architecture: product, user, and order management

Design and development of the NFT&BET platform — https://www.nftandbet.fr, enabling users to buy, sell, and use NFTs as betting tickets for sports events. NFT sales automatically feed a prize pool, which is redistributed to winning participants. 🔹 UX/UI • Smooth interface for NFT buying/selling • Seamless blockchain integration (purchase, resale, results) 🔹 Frontend • Angular (TypeScript) • Dynamic components for NFT and match management 🔹 Backend • Node.js – Custom API for betting logic, user handling, and frontend interactions 🔹 Blockchain • Smart contracts in Solidity: ERC1155, marketplace, game logic • Development with Hardhat & Remix IDE • Deployment on Mumbai (testnet), then Polygon mainnet

Full design and development of a flexible training environment for a neural network specialized in identifying wood species from images (static or video stream). 🔹 Key responsibilities • Managed multiple datasets • Dynamic image preprocessing (rotation, lighting, normalization, etc.) • Implementation and comparison of various CNN architectures (including ResNet and EfficientNet) • Training optimization to fit server constraints (VRAM, time) • Modular system to manage various network/data configurations 🔹 Framework & environment • Python, TensorFlow, CNN architectures • Training run on a Docker server 🔹 Demo application • Web interface allowing users to upload images or video streams for analysis • Explicit results returned by the neural network • Developed with Qt (Python)

Complete design and development of an internal web platform merging an ERP and a PLM system. The platform centralized technical, documentation, and logistics data related to industrial components. Project focused on efficiency, ergonomics, and real-time synchronization with existing systems. 🔹 UX/UI • Designed a dynamic interface with customizable tables, auto-loading, and 3D model/image previews • Advanced flexibility to adapt to different types of equipment and product families • Smooth navigation tailored for industrial users 🔹 Frontend • Django (Python) + AJAX requests for a seamless, no-reload user experience • Integration of dynamic components (filters, pagination, detailed views) 🔹 Backend • Django + SQL • Robust synchronization with ERP and PLM databases • Performance optimization and user rights management

Assisted workers in chemical production.