Sami TAKI

At Farda Tech Consulting, a London-based company, I worked as a freelance AI Engineer across two AI projects. AI Agent SDK — abagent: I developed an agent SDK designed to let developers import their agentic workflows and test how individual nodes behave when parameters change. The SDK enabled teams to measure the impact of these changes on both business and AI metrics, providing a structured way to evaluate workflows and optimize their performance. I delivered a complete milestone following solid software engineering practices around testing, robustness, and maintainability, while working closely with the backend team throughout the project. AI Recruitment Platform — HIME: In parallel, I worked on an AI-powered recruitment mobile application, focusing on the matching and ranking components. I designed the job-candidate matching logic and worked on the ranking model to improve how relevant opportunities were prioritized. I also worked on fine-tuning an AI model for the ranking system, contributing to the development and evaluation of the recommendation pipeline.

Developing an agentic AI system (Project Zephyr) to automate the generation of Software Design Documents (SDD) from aerospace engineering documents (WID, SRD, SAD) using multi-source RAG pipelines and LLMs. - Designed a RAG agent architecture to dynamically orchestrate document reading, cross-referencing and synthesis with full decision traceability. - Orchestrating multi-agent workflows with LangGraph for low-level requirement generation. - Building advanced RAG pipelines with LangChain and Qdrant for semantic retrieval over multimodal industrial data.

IRISA / Multi-Agent System Project Design of an autonomous multi-agent system (reader, analyst, controller, writer) to analyze and synthesize scientific documents. Development of a hybrid RAG pipeline (dense + reranking) for semantic search and retrieval-augmented generation. Collection, cleaning, chunking, and vectorization of heterogeneous data. Evaluation and optimization of generated outputs to ensure coherence and factual consistency. Integration and orchestration with LangGraph, LangChain, and LlamaIndex.

An experience focused on developing an intelligent conversational agent for the HR department, based on RAG (Retrieval-Augmented Generation) technology. Backend development with FastAPI to handle user requests and orchestrate the RAG pipeline. Integration of NLP models for natural language processing and contextual response generation. Design of a document retrieval system based on Qdrant to store and retrieve chunked documents (PDF, Excel, PPT). Extraction and processing of documents using PDF, Excel, and PPT parsers, followed by embedding generation for vector indexing. Implementation of session and document version filtering to personalize responses according to the HR context. Performance and response relevance optimization through the combination of embeddings and OpenAI generation models.

Development of the dashboard module for an LMS (Learning Management System) in React.