AI engineering · autonomous agents · developer tools
Building reliable LLM systems that plan, execute, recover, and validate their work.
Agentic AI Intern at Sofrecom Tunisia · AI Engineering student at ESPRIT · Graduating in 2027
LinkedIn · Email · Flagship project
My main project: a provider-agnostic terminal agent for long-running repository work.
- Plans, edits, executes commands, observes failures, and validates outcomes
- Persists an event ledger with checkpoints, replay, and recovery
- Integrates local LSP and DAP tooling for navigation and debugging
- Supports OpenAI-compatible APIs, Anthropic, Gemini, OpenRouter, Ollama, and LM Studio
- Includes cross-platform tests, security checks, contributor documentation, and architecture notes
Python · Asyncio · LLM agents · LSP · DAP · CI/CD Quick start · Architecture · Demo file
A full-stack application coordinating specialized AI roles across product planning, architecture, infrastructure, security, UI/UX, implementation, and QA. Model outputs are validated as structured artifacts that can be rendered, refined, streamed, persisted, and exported.
Next.js · TypeScript · Zod · Gemini · SSE · Prisma
An educational reference library covering routing, planning, reflection, tool use, memory, RAG, multi-agent coordination, guardrails, evaluation, and other common agent workflow patterns.
Python · LLM architecture · tool use · MCP · evaluation
An academic document-analysis project that extracts ESG-related information from financial documents and checks it against Sustainable Finance Disclosure Regulation requirements.
Python · document processing · regulatory AI · Next.js
At Sofrecom Tunisia, I work on agentic data-reliability workflows: deterministic extraction, semantic alignment, schema-drift detection, controlled ETL repair, sandboxed Python and SQL execution, confidence thresholds, and validation-oriented control flows.
Outside work, I am focused on:
- Reliable autonomous-agent runtimes with explicit state and recovery
- Developer tooling for code navigation, execution, and debugging
- Agent evaluation, observability, and completion integrity
- Practical enterprise AI using retrieval, OCR, and controlled automation
| Area | Technologies |
|---|---|
| Agent systems | Tool calling, MCP, ReAct, LangGraph, LlamaIndex, RAG, evaluation, observability |
| Backend | Python, Asyncio, FastAPI, Redis, PostgreSQL, SQLite, REST, SSE, WebSockets |
| Frontend | TypeScript, React, Next.js, Tailwind CSS |
| Data and AI | Vector search, Pinecone, ChromaDB, FAISS, OCR, anomaly detection, multimodal AI |
| Delivery | GitHub Actions, Docker, automated tests, security scanning |
Languages: Arabic (native), English (fluent), French (fluent).
Open to a 26-week final-year engineering internship starting in early 2027.
AI agents · LLM infrastructure · developer tools · applied software engineering