Jonathan Herrera
Community@jonnabio · Costa Rica
Software Development Director
Agent Skills by Jonathan Herrera
Showing 129 vetted skills indexed across 1 GitHub repositories.
jonnabiodspy
Build AI systems with declarative programming and automated prompt optimization.
jonnabioopenrlhf-training
Train large language models with RLHF algorithms using distributed GPU resources.
jonnabionemo-guardrails
Detect jailbreaks, validate inputs, filter PII, and check toxicity in LLM applications.
jonnabioevaluating-cosmos-policy
Evaluate NVIDIA Cosmos Policy in LIBERO and RoboCasa simulation environments.
jonnabiollamaindex
Build RAG applications with document ingestion, indexing, and querying.
jonnabiodeepspeed
Guide distributed training with DeepSpeed using ZeRO stages, pipeline parallelism, and mixed precision.
jonnabiocrewai-multi-agent
Orchestrates multi-agent AI workflows with role-based collaboration and memory.
jonnabiomoe-training
Train Mixture of Experts models with DeepSpeed and HuggingFace.
jonnabionnsight-remote-interpretability
Interpret and manipulate neural network internals via remote NDIF execution.
jonnabiolangsmith-observability
Trace, evaluate, and monitor LLM operations with the LangSmith framework.
jonnabioaudiocraft-audio-generation
Convert text descriptions into music and sound effects using audiocraft.
jonnabionemo-curator
Processes multimodal LLM training data with GPU-accelerated curation workflows.
jonnabiohuggingface-accelerate
Automate distributed PyTorch training with device placement and mixed precision.
jonnabiochroma
Index and retrieve documents using semantic embeddings and metadata.
jonnabioprompt-guard
Detect prompt injection and jailbreak attempts in LLM applications.
jonnabioevaluating-llms-harness
Evaluate large language models on 60+ academic benchmarks using lm-eval.
jonnabiollama-factory
Automate LLaMA-Factory model fine-tuning with LoRA, unsupervised, or supervised approaches.
jonnabioacademic-plotting
Generate publication-quality diagrams and data figures for machine learning research papers.
jonnabioknowledge-distillation
Compress large language models by distilling knowledge from teacher to student models.
jonnabiosimpo-training
Optimize large language model preferences using the reference-free SimPO method.
jonnabiofine-tuning-openvla-oft
Fine-tunes and evaluates OpenVLA-OFT robot action generation policies on LIBERO and ALOHA environments.
jonnabioslime-rl-training
Optimize LLMs post-training with RL using slime, Megatron-LM, and SGLang.
jonnabioblip-2-vision-language
Generate image captions and answer visual questions with BLIP-2.
jonnabiopytorch-fsdp2
Integrate PyTorch FSDP2 into training scripts for DTensor-based sharding and distributed checkpointing.