Full-Stack Engineering & AI

Achraf Hafsaoui

M.Sc. AI & Machine Learning. Building intelligent systems, scalable backends, and dynamic web applications. Entwicklung intelligenter Systeme, skalierbarer Backends und dynamischer Webanwendungen. Développement de systèmes intelligents, de backends évolutifs et d'applications web dynamiques.

View Work

My engineering philosophy is simple: Build scalable, end-to-end applications powered by intelligent systems.

With a foundation in Computer Science and an M.Sc. in Artificial Intelligence at TU Darmstadt, I focus on the intersection of machine learning and modern full-stack engineering. Whether I am crafting responsive user interfaces, architecting complex databases, or integrating distributed AI agents, my primary drive is shipping robust, production-ready software that delivers real-world value.

Technical Arsenal

AI & Machine Learning

Research & Implementation

Python PyTorch Hugging Face LLMs JAX Reinforcement Learning Prompting Post-processing Methods

Backend & Data

APIs, Databases & Architecture

Go Node.js Django PostgreSQL pgvector Mongoose (MongoDB) REST-APIs WebSockets

Frontend & Mobile

UI/UX & Client Applications

TypeScript JavaScript React React Native Tailwind CSS CSS Vite

DevOps & Tooling

Infrastructure & Testing

Docker GitHub Actions (CI/CD) RabbitMQ Streamlit W&B Jest Postman Agile (Scrum, Jira, Trello)

Featured Engineering Work

03/2026 - Present

AI-Powered CI/CD Debugger

MLOps Architecture

Engineered a distributed, RAG-enabled Al agent that monitors GitHub webhooks, analyzes build logs, and posts code fixes as PR comments.

View Architecture & Code
Python Go PostgreSQL (pgvector) RabbitMQ Docker Streamlit LLMs GitHub Actions

10/2025 - 03/2026

Video Temporal Grounding

Research & Development

Benchmarking and improving multimodal LLM performance in video understanding via systematic prompt engineering and output-processing workflows.

Full research paper currently restricted — Publication pending

"Across Charades-STA, ActivityNet Captions, and QVHighlights, we achieve new state-of-the-art performance among all Training-Free approaches, improving R1@0.5 by +7.5%, +9.2% and +10.9%, respectively—closing the gap to fine-tuned methods."

Research Conclusion

Python PyTorch Hugging Face LLMs (Qwen, GPT) Prompt Engineering Post-processing Techniques

10/2025 - 03/2026

Atari RL: Double Dunk

Reinforcement Learning

Leveraged JAX to train and benchmark comparative RL agents on the Double Dunk Atari environment using pixel and object-based representations.

View Architecture & Code
Python JAX Reinforcement Learning Weights & Biases (W&B) PQN Agents

10/2024 - 03/2025

Bus Monitoring Platform

Agile Development

Developed the MVP as a proof of concept for an early-stage startup, delivering a real-time platform to monitor and manage electric bus fleets.

View Frontend Code View Backend Code

"Overall, I found your work to be structured, reliable, and technically sound. You quickly adapted to new topics and contributed effectively to the team in solving complex tasks."

Dr. Ing. Benjamin Blat Belmonte — Founder, Fenexity

TypeScript React Tailwind CSS Django PostgreSQL Vite REST-APIs (Axios) Postman (tests) Jira (Scrum)

10/2023 - 03/2024

Cross-Platform Collaborative App

Agile Development

Development of a cross-platform app for real-time collaboration featuring chats and whiteboards.

View Architecture & Code
JavaScript React Native Node.js Mongoose (MongoDB) REST-APIs (Axios) WebSockets (Socket.IO) Jest (Tests) CSS Trello (Scrum)

Education

04/2025 - Present

M.Sc. Artificial Intelligence and Machine Learning

Technical University of Darmstadt

10/2021 - 06/2025

B.Sc. Computer Science

Technical University of Darmstadt

Focus: Software Engineering, Web & App Development