anton-abyzov

Hugging Face Ecosystem expert. Covers Transformers (AutoModel, pipeline API), Model Hub, Datasets library, Tokenizers, Accelerate for multi-GPU, PEFT adapters, Evaluate metrics, TGI serving, Spaces deployment, Hub API, Safetensors, Optimum for ONNX/quantization, and Trainer API with custom training loops. Activates for: Hugging Face, HuggingFace, Transformers, Model Hub, TGI, Text Generation Inference, Safetensors, PEFT, Accelerate, Spaces, Datasets, Tokenizers, Optimum.

Mandatory format standard for ALL GitHub issues created by SpecWeave with checkable acceptance criteria and proper metadata. Use when creating GitHub issues, formatting issue content, or ensuring consistent issue structure. Covers user stories, epics, features, and increments.

AWS deep dive expert for CDK, EKS, Lambda, networking, and IAM. Designs production-grade architectures using CDK constructs, serverless patterns, and least-privilege IAM with cross-account access. Generates infrastructure ONE COMPONENT AT A TIME.

MLOps expert for ML infrastructure and production systems. ML pipelines (Kubeflow, Airflow, Prefect), model registry, deployment (Docker, K8s, serverless), monitoring, and CI/CD for ML. Use for automated training, model deployment, or ML infrastructure.

Code navigation with LSP. When user says "find references", "go to definition", "where defined", "show type", "list symbols", "what uses", or "who calls" - use native LSP tools if available, otherwise fall back to specweave lsp commands.

Kafka Connect integration expert. Covers source and sink connectors, JDBC, Elasticsearch, S3, Debezium CDC, SMT (Single Message Transforms), connector configuration, and data pipeline patterns. Activates for kafka connect, connectors, source connector, sink connector, jdbc connector, debezium, smt, data pipeline, cdc.

Expert in Next.js 14+ App Router, Server Components, and Server Actions. Use when building Next.js applications, implementing SSR/SSG, or configuring dynamic routing and data fetching. Covers streaming, caching strategies, middleware, and deployment optimization.

LLM Fine-Tuning expert. Covers LoRA, QLoRA, PEFT, dataset preparation, Hugging Face Trainer/TRL, RLHF, DPO, quantization (GPTQ/AWQ/GGUF), model merging, distributed training with DeepSpeed/FSDP, hardware selection, and production deployment of fine-tuned models. Activates for: fine-tune, fine-tuning, LoRA, QLoRA, PEFT, RLHF, DPO, GPTQ, AWQ, GGUF, LLM training, model training, quantization, DeepSpeed.

RAG (Retrieval Augmented Generation) & Vector Database expert. Covers document loading, chunking strategies, embedding models, vector databases (Pinecone, Weaviate, Qdrant, ChromaDB, pgvector, Milvus), retrieval strategies, query transformation, RAGAS evaluation, multi-modal RAG, and production deployment patterns.

n8n workflow automation with Kafka integration expert. Covers Kafka trigger node, producer node, event-driven workflows, error handling, retries, and no-code/low-code event processing patterns. Activates for n8n kafka, kafka trigger, kafka producer, n8n workflows, event-driven automation, no-code kafka, workflow patterns.

Expert at organizing specs and splitting tasks across multiple GitHub repositories for monorepo, polyrepo, and parent repo architectures. Use when managing specs across multiple repos, coordinating cross-repo work, or allocating tasks to different teams/repositories.

Unit testing and TDD expert for Vitest/Jest. Red-green-refactor workflow, mocking strategies, coverage analysis, and ESM mocking patterns. Use for writing tests, implementing TDD, or setting up test coverage.