Hacker News: maziyar

New comment by maziyar in "Training mRNA Language Models Across 25 Species for $165"

maziyar — Wed, 01 Apr 2026 20:38:33 +0000

full article: https://huggingface.co/blog/OpenMed/training-mrna-models-25-...

Training mRNA Language Models Across 25 Species for $165

maziyar — Wed, 01 Apr 2026 20:38:24 +0000

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.

Comments URL: https://news.ycombinator.com/item?id=47606244

Points: 148

# Comments: 42

New comment by maziyar in "SynthVision: Building a 110K Synthetic Medical VQA Dataset"

maziyar — Mon, 23 Mar 2026 23:29:46 +0000

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.

SynthVision: Building a 110K Synthetic Medical VQA Dataset

maziyar — Mon, 23 Mar 2026 23:29:46 +0000

Article URL: https://huggingface.co/blog/OpenMed/synthvision

Comments URL: https://news.ycombinator.com/item?id=47496552

Points: 3

# Comments: 1

New comment by maziyar in "The ML Engineer's Guide to Protein AI"

maziyar — Thu, 05 Mar 2026 15:33:14 +0000

The 2024 Nobel Prize in Chemistry went to the creators of AlphaFold, a deep learning system that solved a 50-year grand challenge in biology. The architectures behind it (transformers, diffusion models, GNNs) are the same ones you already use. This post maps the protein AI landscape: key architectures, the open-source ecosystem (which has exploded since 2024), and practical tool selection. Part II (coming soon) covers how I built my own end-to-end pipeline.

The ML Engineer's Guide to Protein AI

maziyar — Thu, 05 Mar 2026 15:33:14 +0000

Article URL: https://huggingface.co/blog/MaziyarPanahi/protein-ai-landscape

Comments URL: https://news.ycombinator.com/item?id=47262807

Points: 1

# Comments: 1

New comment by maziyar in "From Golden Gate Bridge to JSON: Why Anthropic's SAE Failed on JSON Output"

maziyar — Wed, 11 Feb 2026 00:28:36 +0000

After six experiments and dozens of failed attempts, I learned something I did not expect: activation steering, the technique Anthropic uses for AI safety, completely fails for one of the most common tasks in production LLM deployments: generating valid JSON.

And I don't mean "fails to help." My steering-only approach achieved 24.4% valid JSON, compared to 86.8% from the completely untrained base model. Steering made the model worse than doing nothing at all.

Here's what I learned, why it matters, and what actually works when you need guaranteed structured outputs from decoder-only language models.

From Golden Gate Bridge to JSON: Why Anthropic's SAE Failed on JSON Output

maziyar — Wed, 11 Feb 2026 00:28:36 +0000

Article URL: https://huggingface.co/blog/MaziyarPanahi/sae-steering-json

Comments URL: https://news.ycombinator.com/item?id=46969144

Points: 2

# Comments: 1

New comment by maziyar in "Trinity large: An open 400B sparse MoE model"

maziyar — Thu, 29 Jan 2026 17:58:04 +0000

i think it's very flattering to have done something with $20m that is so good people think it must have been a $100m!

New comment by maziyar in "Debunking Devin: "First AI Software Engineer" Upwork Lie Exposed [video]"

maziyar — Sat, 13 Apr 2024 06:19:26 +0000

In other words they managed to fake it until they make! Like most visionaries in silicon valley, lie now, tweet about it, prompt it through fake influencers with their mouth open on YouTube, get that VC money without any due diligence, hire smart people and force them to do it!