Frontier AI systems.
Built for production.
AdaLab develops high-performance machine learning architectures. We build consumer-scale multimodal video generation for 220,000+ artists and publish peer-reviewed predictive modeling for clinical medicine.
Plazmapunk — Audio-Reactive Video AI
Our self-developed creative platform turning music tracks into director-level AI videos. Integrates Google Veo 3.1 & LTX 2.5 video diffusion models, multi-scene timeline sequencing, and automatic drum/stem synchronization.
Translational Biomedical Modeling
Rigorous applied ML research in collaboration with university hospitals. We designed predictive neural models for intracranial pressure forecasting in ICU neurocritical care and early postoperative delirium risk stratification.
Plazmapunk
Audio-Reactive AI Video Platform
The next evolution of AI music video generation. Powered by state-of-the-art video diffusion models, multi-scene timeline orchestration, and real-time audio reactivity.
Engineered for Modern Music Creators
Everything built from the ground up for seamless sound-to-visual synchronization.
Three Creation Modes
Tailored workflows for whatever you need to produce:
- • One-Prompt Quick Video: Fast, full-track music video generation in minutes.
- • Scene Editor & AI Director: Multi-scene timeline cuts, granular control & starting images.
- • Spotify Canvas: Seamless 9:16 vertical loops for Spotify, Reels & Shorts.
Next-Gen Video Models
Industry-leading video generation models at your fingertips:
- • Google Veo 3.1: Ultra-photorealistic, high-fidelity diffusion clips with cinematic lighting.
- • LTX 2.5 Video Engine: Native fast video generation with continuous motion up to 24s per scene.
- • Curated Aesthetic Styles: 20+ specialized visual styles from 16mm film to cyberpunk & anime.
Real-Time Audio Reactivity
Music video visuals driven by the actual sonic frequencies of your track:
- • Stem Separation: Isolates drums, bass, instruments, and vocals to steer visual dynamics.
- • Waveform Audio Cuts: Scene transitions lock into musical transients and beat drops automatically.
- • Zero Editing Skills Required: Drag in an MP3 or WAV, and let AdaLab's audio pipeline do the rest.
Join 220,000+ Creators Worldwide
Share your renders, exchange prompt recipes, collaborate with artists, and talk directly to the AdaLab engineering team.
Biomedical AI & Clinical Modeling
Beyond multimodal creator tools, AdaLab's roots lie in rigorous machine learning research for life sciences. We collaborate with leading university clinics to build clinically validated predictive models.
Intracranial Hypertension Prediction in Neurointensive Care
Collaboration with Dr. Nils Schweingruber et al. (UKE Hamburg & Neuro-ICU)
Developed deep recurrent neural network architectures trained on high-frequency physiological time-series data to forecast critical intracranial pressure (ICP) spikes hours before onset, allowing proactive clinical interventions for patients with compromised intracranial space.
- • Dynamic temporal modeling of multi-channel ICU vitals
- • Predicts refractory intracranial hypertension events ahead of acute crisis
- • Peer-reviewed & published in Oxford's flagship neurology journal Brain
Postoperative Delirium Risk Prediction Algorithm (BioCog)
BioCog Consortium & Charité / UMC Utrecht Cohorts
Engineered multivariate statistical and machine learning algorithms evaluating postoperative delirium (POD) risk in elderly surgical patients directly upon emergence from anesthesia, enabling early preventive care and biomarker correlation.
- • Prospective multi-center validation across international cohorts
- • Immediate post-op risk calculation for elderly surgical patients
- • Translational bridge between clinical biomarkers and actionable risk scores
High-Stakes Machine Learning Philosophy
Whether optimizing audio-reactive diffusion for millions of creators or predicting neuro-critical emergencies in intensive care units, our engineering standards are identical: mathematically grounded models, clean evaluation against verified ground truth, and robust zero-downtime deployment.
Meet the Team
We're a small team of ML engineers and developers based in Hamburg. We ship AI products, not slide decks.
Core Team
Anton Wiehe
Co-founder & CTO
Leads machine learning architectures, multimodal diffusion pipelines, and biomedical AI modeling at AdaLab. Co-author of translational research published in Brain and British Journal of Anaesthesia. Graduated top of class in Computer Science at Hamburg University.
Florian Woeste
Co-founder & CEO
Leads strategy, partnerships, and operations across AdaLab and Plazmapunk. Background in finance and enterprise operations before specializing in applied machine learning with an M.Sc. in Machine Learning.
Alex Busch
QA Specialist & Community
Drives product quality assurance, user testing, and creator community engagement. Ensures continuous high reliability for over 220,000 artists creating on Plazmapunk.
Alumni
Pia Čuk
Former Chief Science Officer
Spearheaded early ML research and health-tech modeling. Cognitive Science & ML graduate from Hamburg University.
Artur Galstyan
Former Full-Stack Developer
Built scalable web infrastructure and application interfaces across early AdaLab platforms.
Aida Usmanova
Former ML Developer
Supported machine learning development, data pipelines, and validation.
Bishab Pokharel
Former ML Engineer
Contributed to machine learning pipeline infrastructure and experiment tracking.
Our Partners
Backed by and working with
Get in Touch
Have a project in mind? Let's talk about how we can help.