Data Science & ML Agent Skills

AlphaFold Database Access

Facilitates the retrieval and analysis of over 200 million AI-predicted protein structures from the AlphaFold DB for biological research and drug discovery.

Data Commons Client

Accesses global statistical data from the Data Commons knowledge graph to analyze demographics, economics, health, and environmental trends.

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.

Modernize Scientific Stack

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

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.

Polars High-Performance Data Analysis

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

Protein Assembly

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

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.

Torch Pipeline Parallelism

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

MTEB Retrieval and Ranking

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

PyTorch Model Recovery

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

Scikit-bio Bioinformatics

Performs advanced biological data analysis including sequence manipulation, phylogenetic tree construction, and microbial diversity metrics.

COSMIC Cancer Genomics Access

Accesses the world's largest somatic mutation database for cancer research and precision oncology data retrieval.

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.

SAM Cell Segmentation

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

ReLU Model Extraction

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

DNA Insert Primer Design

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

NCBI Gene Database Tools

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

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.

Distribution Search

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

Protein Assembly

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

G-Code to Text Interpretation

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

Multi-Source Data Merger

Merges heterogeneous data sources into unified datasets with automated field mapping and priority-based conflict resolution.

PyTorch Model CLI

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

MCMC Sampling with Stan

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

Train FastText

Trains and optimizes FastText text classification models while balancing accuracy requirements against model size constraints through systematic hyperparameter tuning and quantization strategies.