https://juanpzuluaga.github.io/ Recent content on Juan Pablo Zuluaga Hugo en-us 2026 Juan Pablo Zuluaga · built with Hugo Thu, 01 Jan 2026 00:00:00 +0000 https://juanpzuluaga.github.io/projects/vllm-omni/ Thu, 01 Jan 2026 00:00:00 +0000 https://juanpzuluaga.github.io/projects/vllm-omni/ Ongoing contributions to vLLM-Omni’s Qwen3-TTS and OmniVoice paths: streaming output, Code2Wav batched decoding, CUDA Graph + torch.compile, voice cloning, and throughput/latency optimization for high-concurrency TTS serving. https://juanpzuluaga.github.io/projects/atco2-corpus/ Sun, 01 Jan 2023 00:00:00 +0000 https://juanpzuluaga.github.io/projects/atco2-corpus/ A multilingual, semi-automatically labeled corpus built to advance ASR and natural language understanding on one of the hardest real-world speech domains. Includes audio, transcripts, speaker role annotations, and a preprocessing pipeline. Used as a benchmark by follow-up work across Europe. https://juanpzuluaga.github.io/projects/wav2vec2-atc/ Sat, 01 Jan 2022 00:00:00 +0000 https://juanpzuluaga.github.io/projects/wav2vec2-atc/ A family of Wav2Vec2 models that achieve 20–40% relative WER reduction on ATC data compared to supervised baselines. Released with training recipes, evaluation scripts, and a Colab notebook for immediate inference. The benchmark paper at SLT 2022 studies self-supervised pretraining behavior under heavy domain shift. https://juanpzuluaga.github.io/projects/bertraffic/ Sat, 01 Jan 2022 00:00:00 +0000 https://juanpzuluaga.github.io/projects/bertraffic/ Most ATC diarization systems rely on audio signals, which are low-quality and short. BERTraffic reframes the problem as text classification: given a transcript, predict speaker turns and whether each turn is a pilot or controller. Beats audio-only baselines by 27% DER. https://juanpzuluaga.github.io/projects/hyperconformer/ Sun, 01 Jan 2023 00:00:00 +0000 https://juanpzuluaga.github.io/projects/hyperconformer/ Attention is the expensive part of Conformer-based ASR models. HyperConformer swaps it for a multi-head HyperMixer, which scales linearly in sequence length rather than quadratically. Same WER as Conformer at a meaningful compute cut. https://juanpzuluaga.github.io/projects/speechbrain/ Mon, 01 Jan 2024 00:00:00 +0000 https://juanpzuluaga.github.io/projects/speechbrain/ SpeechBrain is a PyTorch-based toolkit for speech and language tasks, used by dozens of research groups and startups. The 1.0 release (JMLR 2024) consolidates years of contributions into a stable API with comprehensive recipes for ASR, TTS, speaker recognition, and dialogue understanding. https://juanpzuluaga.github.io/contact/ Mon, 01 Jan 0001 00:00:00 +0000 https://juanpzuluaga.github.io/contact/ How to reach Juan Pablo Zuluaga. https://juanpzuluaga.github.io/cv/ Mon, 01 Jan 0001 00:00:00 +0000 https://juanpzuluaga.github.io/cv/ Curriculum Vitae of Juan Pablo Zuluaga — research, experience, and education.