Senior AI/ML engineer with 15+ years across research, product, and delivery — specializing in LLM integration, RAG pipelines, and multi-agent conversational systems. PhD in Computer Science. Full-stack capable, so AI features ship end-to-end, not just as prototypes.
From model/API selection to production deployment — I deliver AI features that are grounded, monitored, and maintainable, not just demos.
Multi-turn dialogue, intent handling, tool use, and multi-agent orchestration built on GPT, BERT, Codex, DialoGPT, and DALL-E-based flows.
Retrieval-Augmented Generation over company documents — vector stores, embeddings, and retrieval layers for accurate, source-grounded responses.
12+ specialized agent types shipped: customer support, sales, booking, e-commerce, HR, education, healthcare, legal/financial, and RAG knowledge bases.
Real-time dashboards for inference, training convergence, GPU utilization, and vector DB health — production-grade observability for AI systems.
React/TypeScript front-ends, Node.js/Express APIs, auth and Stripe billing — so AI features ship as complete products, not isolated scripts.
Speech-to-text and voice-driven conversational interfaces, plus generative image flows for creative and product use cases.
Live, runnable demonstrations of AI/ML and full-stack delivery — not slides. Full portfolio at clavie.fr.
Support, sales, booking, e-commerce, HR, education, healthcare, legal/financial, RAG knowledge base, social, voice, and autonomous multi-agent — each runnable and documented.
Real-time metrics for inference, training, and database health — total requests, latency, active GPU clusters, vector DB size, and live training convergence.
SaaS task manager (auth, Stripe, roles), e-commerce store, real-time chat with WebSockets, and a Notion-style notes app — built to host or integrate AI features.
Generative AI for workout and nutrition; LLM-based coaching (Swift/Kotlin, TensorFlow Lite, GPT-3).
Text/sketch to images (React Native, PyTorch, DALL-E 2 API).
Personalized recommendations and price comparison (Flutter, TensorFlow, BERT).
Personalized lessons, grammar, pronunciation (Swift/Kotlin, TensorFlow, GPT-3).
Empathetic conversations, mindfulness, wearable integration (React Native, TensorFlow, DialoGPT).
Dynamically generated world, characters, storylines (Unity, C#, GPT-2).
Personalized itineraries, travel guides, Q&A (Flutter, TensorFlow, GPT-3).
Natural language to code snippets/programs (Swift/Kotlin, TensorFlow, Codex).
Customized learning materials and assessments (React Native, TensorFlow, GPT-3).
Scheduling, reminders, information retrieval, smart home (Flutter, TensorFlow, Dialogflow).
General web/app development work — e-commerce, SaaS, corporate sites, dashboards, and landing pages.
→ webdev.html 🤖 Live PortfolioA separate portfolio dedicated to custom AI chatbots, conversational agents, and automation built for real businesses.
→ products.clavie.frFrom semiconductor R&D to release management to edge-computing research — a systems background that now underpins production AI delivery.
Principal contributor to a heterogeneous computing design (VMs, micro-VMs, storage, unikernels) for optimized and secure workloads. Built scalable storage infrastructure (MongoDB); outcomes directly relevant to edge AI and low-latency inference.
Led software release strategy and execution with cross-functional teams; automated build processes and deployments from development to production — skills transferable to AI/ML pipeline and model release workflows.
Managed release planning and scheduling for globally distributed engineering teams; rigorous version control and automated builds, delivered on time through remote collaboration across time zones.
In-house rapid prototyping platform with test-driven development and verification; Linux administration and cross-team collaboration.
Synthesis and verification of imaging IPs; bit-true functional verification and stress testing (e.g. H.264 SVC); non-regression and validation scripting.
A stack spanning model integration, data, back-end, and front-end — chosen for production reliability.
Peer-reviewed work in systems and edge computing — applied to latency, security, and scalability of AI services.
Enhancing Edge Computing with Unikernels in 6G Networks — S. Yazdani, N. Ramzan, P. Olivier. Unikernel-based edge computing for improved performance, security, and low latency; relevant to edge AI and inference deployment.
Compiler and System Techniques for SoC Distributed Reconfigurable Accelerators — S. Yazdani et al. High-level programming and tooling for heterogeneous compute.
Coordinated Concurrent Memory Accesses on a Reconfigurable Multimedia Processor — S. Yazdani, J. Cambonie, B. Pottier. Programming model and scheduling for accelerators (65nm STMicroelectronics).
Programming Reconfigurable Decoupled Application Control Accelerator for Mobile Systems — S. Yazdani, J. Cambonie, B. Pottier. Multimedia accelerator with scratchpad memory; 65nm, 200 MHz.
Université de Bretagne Occidentale (UBO), France. Thesis on coordinated memory access and reconfigurable multimedia processors; published at ReCoSoC, SLAP, Springer.
Signals & Circuits — UBO, France.
NED University of Engineering & Technology, Karachi, Pakistan.
Endorsements submitted by people I've worked with — verified and reviewed before appearing here.
Received an invite from me? Use the same link to submit yours.
Whether it's LLM integration, a RAG pipeline, a conversational agent, or a full-stack AI product — let's talk about your use case.
Start a Project →Open to freelance projects, contract work, and long-term collaborations across time zones. I'll get back to you within 24 hours.