Youssef Mehili
Software Engineer (Backend & AI)
About
I am a software engineer and AI enthusiast based in Lyon, France, currently pursuing a Master's degree at EPITECH. With a deep passion for bridging complex backend architectures with cutting-edge artificial intelligence, I strive to build systems that are not only robust but also intelligent.
My journey spans from low-level C++ optimization for critical infrastructure at the Ministry of Transport to architecting scalable microservices and AI integrations as a Lead Developer. I enjoy tackling challenging problems, whether it's developing neuromorphic algorithms, building custom LLMs from scratch, or ensuring national safety compliance through software.
Beyond technical expertise, my experience in sales leadership has given me a unique perspective on product value and business strategy. I am driven by curiosity and a commitment to innovation, constantly seeking to push the boundaries of what's possible in software engineering.
Experience
Serfim TIC
Mar 2026 – Present- Accelerated secure enterprise-wide AI adoption across 3+ internal workflows through privacy-preserving AI integrations using autonomous agents and production-safe data protection patterns.
- Reduced manual processing time by an estimated 30–40% through autonomous workflow automation using AI agents integrated into existing production systems.
- Improved application reliability across 40+ tickets through bug fixes, performance optimizations, and feature maintenance using Vue.js, Nest.js, and Java.
Tunnel Study Center (CETU) — Ministry of Transport
Jul 2024 – Mar 2026- Supported national tunnel safety compliance through the redesign of critical lighting sizing software using C++ and Python.
- Improved calculation and rendering performance by an estimated 35–50% through optimized visibility algorithms using physics-based light modeling techniques.
- Reduced engineering iteration time by an estimated 30–40% through backend modernization and UX improvements using Python, C++, and simulation workflow optimization.
Junior Conseil Taker
Jan 2025 – Present- Improved delivery scalability across 5+ client projects through robust API and microservice architectures using Nest.js, Node.js, and FastAPI.
- Reduced manual business workflow effort by an estimated 25–35% through custom AI automation systems using Python, PyTorch, LLMs, and computer vision models.
- Increased delivery reliability across development teams through code reviews, QA protocols, and CI/CD workflows using Agile project management practices.
Junior Conseil Taker
Sep 2025 – Present- Secured €220k in annual turnover through strategic client acquisition and retention using data-driven sales pipelines and CRM workflows.
- Improved team execution across 20 sales representatives through structured KPI tracking and weekly coaching using performance dashboards and sales routines.
- Increased deal conversion consistency through standardized negotiation and contract workflows using CRM follow-ups and improved commercial processes.
Junior Conseil Taker
May 2025 – Present- Expanded the qualified sales pipeline through 300+ targeted outreach actions using email, phone, LinkedIn, and CRM automation tools.
- Accelerated client acquisition through high-potential lead qualification using consultative selling techniques and strategic appointment scheduling.
- Improved lead conversion rates through pitch optimization and outreach experimentation using A/B testing and market trend analysis.
PoC Innovation
Sep 2024 – Jul 2025- Advanced neuromorphic AI research through 3+ bio-inspired prototypes using PyTorch and spiking neural network architectures.
- Validated theoretical models through end-to-end proofs of concept using physical or simulated hardware and reproducible experimentation pipelines.
- Accelerated student knowledge transfer through practical R&D projects and open-source learning material using recent academic research, Python, and PyTorch.
Projects
Supported safety compliance for French tunnel infrastructure through high-precision lighting sizing software using C++ and Python, reducing engineering iteration time by an estimated 30–40%.
Public reference + pictures available.
Reduced manual suspicious-check verification effort through a fraud detection SaaS with forensic analysis capabilities using Next.js, FastAPI, AWS, TailwindCSS, and shadcn/ui.
Enabled real-time multiplayer gameplay through a scalable C++ game engine ecosystem with networking, scripting, and asset management using hexagonal architecture.
Streamlined cross-platform task automation through a trigger-action mobile and web system using FastAPI, Python, Next.js, Tailwind, and React Native.
E-Commerce for West Africa
Enabled scalable online retail operations for West Africa through an Amazon-style marketplace using Next.js, React, Node.js, Prisma, SQL, Tailwind, and Stripe.
NDA
Improved rendering scalability for photorealistic scenes through a distributed ray tracer with acceleration structures using C++.
Deepened practical understanding of LLM architecture through a GPT-2 implementation and training pipeline from scratch using Python and PyTorch.
Simulated complex autonomous team behavior through a distributed multiplayer strategy game using a C++ server and Python AI agents.
Company Document RAG
Reduced internal knowledge search time by an estimated 40–60% through a grounded RAG search engine using LangChain, Ollama, and Python.
NDA
Insurance Chatbot
Reduced repetitive customer support workload by an estimated 25–35% through an intelligent insurance chatbot using the OpenAI API, React, and Next.js.
NDA
Accelerated experimentation across multiple vision and reinforcement learning prototypes through reusable training and evaluation pipelines using Python and PyTorch.
Biotech Finance AI Newspaper
Reduced manual finance article preparation time by an estimated 30–50% through an AI-assisted recommendation and writing platform using React Native, Node.js, and the OpenAI API.
Link/app coming soon.
Robot Dog
Advanced autonomous quadruped movement through control and perception experiments using reinforcement learning, PyTorch, and simulation pipelines.
Videos coming soon.