AI Skills for Data Scientist

Conducts a comprehensive audit of agent and LLM applications, identifying failures and providing code-first fixes for improved performance.

Enhances output quality by enforcing fact-checking before code edits, ensuring thorough investigation and context awareness.

Designs composable recommendation and ranking pipelines using a six-stage framework for personalized content delivery.

Facilitates structured self-debugging for AI agents, enhancing their ability to diagnose and recover from failures systematically.

Enhances security for autonomous trading agents by implementing robust patterns against financial threats and transaction vulnerabilities.

Facilitates quick genomic database queries and bioinformatics evidence logging using the gget CLI and Python workflow.

Implements eval-driven development for Claude Code sessions, enhancing AI reliability through structured evaluation frameworks.

Facilitates production-ready machine learning workflows with data contracts, reproducible training, and robust model evaluation and monitoring.

Provides best practices and patterns for building efficient and reproducible deep learning applications using PyTorch.

Automatically extracts reusable patterns from Claude Code sessions for future use, enhancing learning and efficiency.

Enhances multi-agent workflows by refining context retrieval through iterative search patterns, improving task execution efficiency.

Optimizes LLM API costs with model routing, budget tracking, retry logic, and prompt caching for efficient usage.

Compares coding agents like Claude Code and Aider on custom tasks, measuring pass rates, costs, time, and consistency.

Provides best practices and idioms for building robust and maintainable Python applications, focusing on readability and type hints.

Enhances transformer models by optimizing attention mechanisms for significant speed and memory efficiency improvements.

Facilitates scalable and efficient training of PyTorch models with minimal boilerplate and built-in best practices.

Guides users in training Sparse Autoencoders to analyze neural network activations and discover interpretable features.

Guides LLM post-training with RL using slime, integrating Megatron-LM for custom workflows and high-throughput data generation.

Facilitates distributed LLM pretraining in PyTorch using TorchTitan, optimizing for performance across multiple GPUs.

Facilitates the end-to-end process of writing ML research papers for top conferences, from design to submission.