Training mRNA Language Models Across 25 Species for $165
•
6
AI & ML interests
Health x AI
Recent Activity
View all activity
Articles
MaziyarPanahiÂ
updated a
model 1 day ago
MaziyarPanahiÂ
published a
model 1 day ago
MaziyarPanahiÂ
posted an update 1 day ago
Post
152
Training mRNA Language Models Across 25 Species for $165
We built an end-to-end protein AI pipeline covering structure prediction, sequence design, and codon optimization. After comparing multiple transformer architectures for codon-level language modeling, CodonRoBERTa-large-v2 emerged as the clear winner with a perplexity of 4.10 and a Spearman CAI correlation of 0.40, significantly outperforming ModernBERT. We then scaled to 25 species, trained 4 production models in 55 GPU-hours, and built a species-conditioned system that no other open-source project offers. Complete results, architectural decisions, and runnable code below.
https://huggingface.co/blog/OpenMed/training-mrna-models-25-species
We built an end-to-end protein AI pipeline covering structure prediction, sequence design, and codon optimization. After comparing multiple transformer architectures for codon-level language modeling, CodonRoBERTa-large-v2 emerged as the clear winner with a perplexity of 4.10 and a Spearman CAI correlation of 0.40, significantly outperforming ModernBERT. We then scaled to 25 species, trained 4 production models in 55 GPU-hours, and built a species-conditioned system that no other open-source project offers. Complete results, architectural decisions, and runnable code below.
https://huggingface.co/blog/OpenMed/training-mrna-models-25-species
MaziyarPanahiÂ
published an article 1 day ago
Article
Training mRNA Language Models Across 25 Species for $165
•
6
MaziyarPanahiÂ
posted an update 8 days ago
Post
2104
We annotated 119K medical images with two frontier VLMs (Qwen 3.5, Kimi K2.5), cross-validated at 93% agreement, and produced 110K training records, all for under $500. Fine-tuning 3 small models (2-3B params) improved all benchmarks: best model reaches +15.0% average exact match.
Everything is open-sourced: datasets, adapters, and code.
https://huggingface.co/blog/OpenMed/synthvision
Everything is open-sourced: datasets, adapters, and code.
https://huggingface.co/blog/OpenMed/synthvision
- 2 replies
MaziyarPanahiÂ
published an article 9 days ago
Article
SynthVision: Building a 110K Synthetic Medical VQA Dataset with Cross-Model Validation
•
15
MaziyarPanahiÂ
published 3
models 9 days ago
MaziyarPanahiÂ
published 5
datasets 9 days ago
OpenMed/synthvision-seeds
Viewer • Updated • 119k • 245 • 2
OpenMed/synthvision-annotated-qwen
Viewer • Updated • 59.5k • 130 • 1
OpenMed/synthvision-annotated-kimi
Viewer • Updated • 59.5k • 131 • 1
OpenMed/synthvision-validated-qwen-by-kimi
Viewer • Updated • 115k • 154 • 1
OpenMed/synthvision-validated-kimi-by-qwen
Viewer • Updated • 115k • 185 • 2
MaziyarPanahiÂ
updated 3
models 9 days ago
MaziyarPanahiÂ
updated 3
datasets 9 days ago