Projects

DeepFense, Modular Deepfake Audio Detection Framework

An open-source, configuration-driven framework I co-built for deepfake audio detection research. Mix and match SSL frontends, classifier backends, losses, and augmentations entirely via YAML, without touching code.

• Plug-and-play registry: Wav2Vec 2.0, WavLM, HuBERT, MERT, EAT frontends and AASIST, ECAPA-TDNN, Nes2Net, RawNet2 backends.
• 455+ pretrained models and 12 benchmark datasets (ASVSpoof19, CompSpoof, DECRO, SONICS) on HuggingFace.
• Built-in augmentations (RawBoost, RIR, codec, noise) and standardized metrics (EER, minDCF, actDCF, F1).

News-Polygraph

A collaborative German research initiative building a multimodal platform for detecting and analyzing disinformation across speech, image, and text. My contribution is on the speech and audio side: robust generalizable deepfake detection, anti-spoofing, and interpretability, feeding into the DeepFense framework above.

Mamba, SSL Interpretability, and Spectral Fusion

A connected research line on making speech anti-spoofing systems both stronger and more efficient:

• Parameter-Efficient Multi-Scale Adapter (ICASSP 2026): adapter-style fine-tuning of SSL backbones for synthetic-speech detection, with sharp efficiency gains.
• BiCrossMamba-ST (Interspeech 2025): among the first successful applications of Mamba state-space models to speech anti-spoofing. SOTA on ASVSpoof LA21 and DF21 with 28% fewer parameters than Transformer baselines.
• Two Views, One Truth (WASPAA 2025): hybrid fusion of SSL embeddings with handcrafted spectral descriptors (MFCC, LFCC, CQCC) for cross-domain robustness.
• Layer-wise SSL analysis (NAACL Findings 2025): first large-scale study showing that lower layers of speech SSL models are sufficient for artifact detection, enabling ~3x faster inference.

AtlasIA, Open-Source AI for Morocco

Leading the Speech Team in an open-source community building AI grounded in Moroccan linguistic identity:

• MoulSot (2026): curated 80 hours of high-quality Moroccan Darija speech distilled from 1,500 hours of YouTube content through a 9-stage pipeline (Silero VAD, SQUIM, Audiobox Aesthetics, DNS64 denoising, pyannote diarization, Argilla human-in-the-loop, Gemini 2.5 Pro transcription), then fine-tuned Qwen3-ASR-1.7B on top. Released as atlasia/MoulSot-Full and atlasia/moulsot.v0.3.
• Building the first natural Text-to-Speech system and tokenizer for Moroccan Darija.
• Mentoring HackAI Morocco teams on forced alignment and low-resource ASR.

Iqra'Eval, Quranic Pronunciation Assessment

A multi-institutional initiative to benchmark Arabic pronunciation assessment, using Quranic recitation as the case study:

• Founded and ran the Iqra'Eval Shared Task at ArabicNLP 2025, establishing the first standard benchmark for Quranic recitation analysis.
• Co-built QuranMB, the first expert-annotated dataset for Quranic mispronunciation (~2 h), with baselines, originally as lead of a 15-researcher team at the SDAIA Winter School.
• Next iteration accepted as an Interspeech 2026 Challenge (IQRA 2026).

QVoice and the Arabic Speech Stack

Led QVoice, the first end-to-end mispronunciation detection system for Modern Standard Arabic, and contributed a series of SOTA results on L2 speech assessment and multilingual ASR:

• SpeechBlender (SLaTE 2023), fine-grained augmentation for mispronunciation data scarcity.
• L1-aware Multilingual MDD (ICASSP 2024), incorporating native-language priors into multilingual mispronunciation detection.
• Beyond Orthography (ACL 2024), recovering short vowels and dialectal sounds in Arabic with limited data.
• AraVoiceL2, non-native Arabic speech dataset for phoneme-level evaluation.

Fanar & MorphBPE

Contributed to Fanar, Qatar's Arabic-centric multimodal LLM, and designed MorphBPE, a morphology-aware tokenizer that integrates linguistic structure into subword segmentation. The work shows how foundation-model design choices can be tuned to morphologically rich languages without sacrificing efficiency.

• MorphBPE improves fertility rates and downstream generation quality for morphologically rich languages, with negligible overhead vs. plain BPE.
• Helped run large-scale data filtering and training pipelines (300M to 3B parameters) with MosaicML's LLM-Foundry.

Gretchen AI, Research & Engineering Lead

Co-leading R&D for generative-media detection at a Berlin-based AI startup. Built a proprietary image deepfake detector that ranked #1 on the Deepfake-Eval 2024 benchmark (~83% accuracy), and designed the scalable inference pipeline used in production.