Nolla Health worked with TrainLoop to post-train NVIDIA Nemotron 3.5 Super VL on 32 NVIDIA B200 GPUs provided by Nebius. The NollaMD project trains models to interpret clinical images, gather patient histories and produce clinical assessments. Training used NVIDIA NeMo RL and Group Relative Policy Optimization, with an initial visual-question stage followed by simulated patient consultations.
Nolla reported that accuracy on its 500-question internal visual evaluation rose from 35.0% to 66.8%. At a conversational checkpoint, completed consultations increased from 50 to all 75 held-out simulated encounters, and top-diagnosis accuracy rose from 48% to 64%. These results cover internal evaluations and simulated encounters. Nolla identified evaluation in real encounters and better recognition of urgent cases as next steps.
Nolla Health reports using 32 Nebius-provided B200 GPUs with TrainLoop to post-train Nemotron on clinical visual questions and simulated patient consultations. It reports internal evaluation improvements and identifies real-encounter evaluation as a next step.
Training NVIDIA Nemotron for Clinical CarePublished 15 Oct 2026 · Nolla Health · Web pageNolla Health reports using 32 Nebius-provided B200 GPUs with TrainLoop to post-train Nemotron on clinical visual questions and simulated patient consultations. It reports internal evaluation improvements and identifies real-encounter evaluation as a next step.
Nolla Health, operated by Magic Health, Inc., develops consumer healthcare software combining AI tools with access to licensed clinicians. Its Nolla Derm app offers skin assessments, personalized skincare and prescription treatment, with tools for tracking progress over time. The platform serves people seeking remote care for acne and other skin concerns. Nolla also offers a beta health assistant for symptoms and care navigation, while its planned Nolla Women product focuses on women's health needs.
Nolla announces NVIDIA launch partnership and clinical model results
Nolla identified itself as an NVIDIA launch partner and described training Nemotron 3.5 Super VL with TrainLoop for clinical image interpretation and simulated consultations. It reported visual accuracy rising from 35.0% to 66.8% on its internal evaluation, with patient-care validation still pending.
Nolla announces women's health app planned for early 2027
Nolla outlined Nolla Women, an app being developed for personalized check-ins, symptom and cycle tracking, and care guided by licensed medical providers. The company expects it to go live in early 2027.
Nolla introduced forecasting that uses consecutive facial scans to estimate acne flare-up risk over the next three days. The feature became available in Nolla Derm on the announcement date.
Healthcare technology company offering a health assistant, dermatology care and access to licensed clinicians. · Health
How they use Nebius AI Cloud
Nolla Health worked with TrainLoop to post-train NVIDIA Nemotron 3.5 Super VL on 32 NVIDIA B200 GPUs provided by Nebius. The NollaMD project trains models to interpret clinical images, gather patient histories and produce clinical assessments. Training used NVIDIA NeMo RL and Group Relative Policy Optimization, with an initial visual-question stage followed by simulated patient consultations.
Nolla reported that accuracy on its 500-question internal visual evaluation rose from 35.0% to 66.8%. At a conversational checkpoint, completed consultations increased from 50 to all 75 held-out simulated encounters, and top-diagnosis accuracy rose from 48% to 64%. These results cover internal evaluations and simulated encounters. Nolla identified evaluation in real encounters and better recognition of urgent cases as next steps.
Nolla Health reports using 32 Nebius-provided B200 GPUs with TrainLoop to post-train Nemotron on clinical visual questions and simulated patient consultations. It reports internal evaluation improvements and identifies real-encounter evaluation as a next step.
Training NVIDIA Nemotron for Clinical CarePublished 15 Oct 2026 · Nolla Health · Web pageNolla Health reports using 32 Nebius-provided B200 GPUs with TrainLoop to post-train Nemotron on clinical visual questions and simulated patient consultations. It reports internal evaluation improvements and identifies real-encounter evaluation as a next step.
Nolla Health, operated by Magic Health, Inc., develops consumer healthcare software combining AI tools with access to licensed clinicians. Its Nolla Derm app offers skin assessments, personalized skincare and prescription treatment, with tools for tracking progress over time. The platform serves people seeking remote care for acne and other skin concerns. Nolla also offers a beta health assistant for symptoms and care navigation, while its planned Nolla Women product focuses on women's health needs.
Nolla announces NVIDIA launch partnership and clinical model results
Nolla identified itself as an NVIDIA launch partner and described training Nemotron 3.5 Super VL with TrainLoop for clinical image interpretation and simulated consultations. It reported visual accuracy rising from 35.0% to 66.8% on its internal evaluation, with patient-care validation still pending.
Nolla announces women's health app planned for early 2027
Nolla outlined Nolla Women, an app being developed for personalized check-ins, symptom and cycle tracking, and care guided by licensed medical providers. The company expects it to go live in early 2027.
Nolla introduced forecasting that uses consecutive facial scans to estimate acne flare-up risk over the next three days. The feature became available in Nolla Derm on the announcement date.
