Data Science & ML Agent Skills

NCBI Gene Database Tools

Queries and retrieves genomic data from NCBI Gene databases using E-utilities and the modern Datasets API.

Statsmodels Statistical Analysis

Performs rigorous statistical modeling, econometric analysis, and hypothesis testing using Python's statsmodels library.

Transformers

Integrates state-of-the-art machine learning models for NLP, computer vision, and audio tasks using the Hugging Face ecosystem.

Modernize Scientific Stack

Migrates legacy Python 2 scientific computing code to Python 3 using modern libraries like pandas, numpy, and pathlib.

LLM Inference Batching Scheduler

Optimizes LLM inference request grouping and scheduling to minimize operational costs while satisfying latency and padding constraints.

ARC-AGI Git Merge & Pattern Tasks

Facilitates solving complex pattern recognition tasks by combining git workflow management with mathematical grid transformation analysis and implementation.

PyTorch Model Recovery

Reconstructs PyTorch model architectures from weight files and state dictionaries by analyzing tensor shapes and naming patterns.

Raman Spectroscopy Peak Fitting

Analyzes and fits peaks in Raman spectroscopy data using physically-constrained models like Lorentzian, Gaussian, and Voigt functions.

Data Resharding and Partitioning

Reorganizes large-scale datasets into hierarchical directory structures while enforcing strict file size and item count constraints.

Torch Pipeline Parallelism

Implements PyTorch pipeline parallelism to distribute large language model training across multiple GPUs using All-Forward-All-Backward (AFAB) scheduling.

Protein Assembly

Designs and optimizes multi-component fusion protein sequences for FRET biosensors and gene synthesis.

Protein Assembly

Designs optimized DNA gBlock sequences for fusion proteins by combining sequences from multiple databases with precise linker and codon constraints.

Modernize Scientific Stack

Upgrades legacy Python 2 scientific computing code and analysis pipelines to modern Python 3 standards using contemporary libraries like NumPy and pandas.

MTEB Retrieval and Ranking

Optimizes semantic similarity retrieval tasks through expert guidance on document preprocessing, embedding model selection, and similarity ranking.

Scientific Hypothesis Generator

Generates testable, evidence-based scientific hypotheses and experimental designs from observations or literature.

Histolab Digital Pathology

Processes whole slide images (WSI) for digital pathology by automating tissue detection, tile extraction, and preprocessing for computational pipelines.

ReLU Model Extraction

Extracts internal weight matrices and biases from black-box ReLU neural networks using input-output query strategies and functional equivalence testing.

Raman Fitting

Provides specialized guidance and implementation patterns for analyzing and curve-fitting peaks in Raman spectroscopy data.

PyTorch Model Recovery

Reconstructs PyTorch model architectures from saved state dictionaries, enables selective layer fine-tuning, and facilitates TorchScript conversion for deployment.

Distribution Search

Finds probability distributions that satisfy specific statistical constraints like KL divergence targets through mathematical analysis and optimized parameterization.

SAM Cell Segmentation

Implements SAM-based biological image segmentation pipelines, converting binary masks to polygon coordinates for microscopy data processing.

Polars High-Performance Data Analysis

Optimizes data processing workflows using the high-performance Polars DataFrame library and expression API.

DNA Assembly

Designs optimized primers and validates multi-fragment DNA assembly workflows for Golden Gate and Type IIS cloning techniques.

HuggingFace Model Inference

Deploys pre-trained HuggingFace Transformer models as robust REST API inference services using Flask or FastAPI.

RStan to PyStan Translation

Translates Bayesian inference workflows and Stan model implementations from RStan to PyStan with high precision and numerical stability.

DNA Insert Primer Design

Designs specialized primers for inserting DNA sequences into circular plasmids using Q5 site-directed mutagenesis and inverse PCR techniques.

MCMC Sampling with Stan

Implements Bayesian Markov Chain Monte Carlo sampling workflows using RStan for complex hierarchical modeling and statistical inference.

G-Code to Text Interpretation

Decodes and interprets text content from G-code files by analyzing geometric toolpath data and coordinate patterns.

MuJoCo MJCF Model Tuning

Optimizes MuJoCo MJCF model files for simulation performance while maintaining numerical accuracy and physical correctness.

PyTorch Model CLI

Converts PyTorch neural networks into standalone C/C++ command-line tools by extracting weights and reimplementing inference without Python dependencies.