Bachelor Thesis

Vehicle detection in urban areas faces accuracy issues due to occlusions; comparing two machine learning models aimed at inpainting occluded vehicles to improves detection and data reliability.

A secure location-based service using attribute-based credentials and cryptographic pairings with Petrelic. Built and deployed isolated client-server architecture using Docker, with privacy protection via Tor integration and tested robustness using pytest.

Built a Secure Multi-party Computation (SMC) system in Python, implementing secret sharing, Beaver triplet generation, and arithmetic circuit evaluation for private collaborative computation. Designed expression parsing and integrated robust inter-party communication with tested architecture using pytest and Flask.

Implemented low-level CPU and DRAM attacks such as AnC, Meltdown, RetBleed, and Rowhammer, targeting memory isolation, speculative execution, and physical reliability. Developed real-world exploits requiring deep expertise in computer architecture, systems programming, side-channel analysis, and hardware reverse engineering.

Built a fully functional ACMEv2 client ecosystem supporting dns-01 and http-01 challenges, wildcard domains, and automated certificate revocation using a custom DNS, HTTP, and HTTPS stack. Implemented JOSE-compliant request signing, X.509 certificate handling, and RFC8555-compliant state machines, integrated with Pebble for end-to-end testing.

A reinforcement learning-based system to detoxify hate speech while preserving semantic meaning, combining a fine-tuned BART model with a custom reward function based on BERTScore and a toxicity classifier. Trained under resource-constrained conditions, Developed skills in reinforcement learning, model optimization, and meaning metrics.

Simple educational operating system inspired by MIT’s JOS, implementing core kernel components including process isolation, virtual memory, preemptive multitasking, file systems, and basic system call handling. Gained experience in x86 architecture, low-level C, memory management, concurrency primitives, and debugging with gdb in a bare-metal environment.

BeGreen

BeGreen is an Android app in Kotlin which gamifies waste collection, offering rewards and connecting users with recycling resources to inspire a cleaner planet. Made in Kotlin

Learning The Secrets Of Data

Identify trends shared among successful music composers and compositions among API asynchonous call and Data analysis

BowSteroid

An NDS game in C where you are a spaceship lost in space and some asteroids appears randomly. You need to survive by avoid them or destroying them.

BIO-482 Miniproject

Data analysis miniproject of the course Neuroscience: cellular and circuit mechanisms

Learning The Secrets Of Data Website

The website of the LSD project

CKVS

A key-value store project completely programmed in C oriented system

tCHu

An interactive multiplayer board game about train rails in switzerland completely programmed in Java

CPU in ASM/VHDL

We simulated the mechanism of a basic CPU on a FPGA mainly in ASM and VHDL