Generates code to build a sentence similarity detection model by extracting BERT embeddings and feeding them into a Bi-LSTM network using TensorFlow and Hugging Face Transformers.
This skill is used to transfer attributes from a source character description to a target character description. It ensures the target character retains their unique integrity without degradation, removes direct references to the source character, and enhances the resulting text.
Converts SPICE-like circuit netlists into NetworkX MultiGraphs with randomized parameters, specific node/edge feature schemas, and multi-edge handling for Graph Neural Network Reinforcement Learning models.
Rewords or paraphrases performance criteria for training contexts and adds concrete, observable workplace examples to strengthen assessment context.
Executes binary classification using DNN and CNN models, with and without CHAID feature selection, using a rolling time-series training window. Handles missing data via mean imputation and outputs a CSV with appended prediction columns.
Recommends top electric vehicles based on user-defined technical constraints (features, range, body type, brand) and calculates a weighted percentage match score for ranking.
Extracts embedding vectors from the layer immediately preceding the Softmax layer of a pre-trained model (e.g., Inception-V3, ResNet50) and saves them in a dictionary where the key is the embedding vector and the value is the corresponding label.
Эксперт по Julia для реализации функций условного изменения размерности данных (reshape) и последовательного повторения элементов матрицы с соблюдением строгих типизаций.
Generates a Python script using Keras and scikit-learn to perform grid search hyperparameter tuning. The script tests different model architectures, layers, GRU units, optimizers, and regularizations, running for approximately 50 epochs and printing statistics to the console.
Generates MATLAB code to perform time series forecasting by testing multiple ARIMA models, selecting the best one based on AIC, and plotting the forecast.
Implements an iterative LDR decomposition of a real matrix X using QR factorization, following specific initialization, update rules, and convergence criteria.
Generate executable MATLAB code to simulate traffic flow on a single lane using specified car-following models (e.g., IDM, Gipps, Krauss). The code must handle cars of varying dimensions, ensure collision avoidance, animate the results, and be free of syntax errors or index out-of-bounds issues.
Generates a Python function to train a CatBoost model on a DataFrame and plot feature importances. The plot must display feature names on the y-axis and assign a unique color to each feature bar.
A skill to conditionally update a target column in a pandas DataFrame based on a reference column and specific string matching rules, handling nulls and type errors.
Converts PyTorch training loop code from using CrossEntropyLoss to MSELoss, specifically updating the accuracy calculation logic from argmax-based comparison to rounding-based comparison to handle regression outputs.
A skill to classify samples from two lists of variable-length 1D arrays using PyTorch. It includes specific preprocessing rules for truncation and zero-padding, handles imbalanced datasets via resampling, and enforces specific evaluation metrics including Precision, Recall, F1-score, and Confusion Matrix.
Load flattened image data from a CSV file with specific schema (label + pixel columns), create train/validation/test data loaders, and train a retrained VGG classification network.
Roleplay as a Remote Viewing AI trained on coordinate/target pairings to describe targets, answer questions, and provide recommendations.
Use when running Sprint Refinement sessions with SFDIPOT product factors, generating BDD scenarios, or validating requirements in the QCSD Refinement phase.
Predicts defect-prone code using change frequency, complexity metrics, and historical bug patterns. Use when predicting defects before they escape, analyzing root causes of test failures, learning from past defect patterns, or implementing proactive quality management.
Optimizes QE agent performance through transfer learning, hyperparameter tuning, and pattern distillation across test domains. Use when improving agent accuracy, applying learned patterns to new projects, tuning quality thresholds, or implementing continuous improvement loops for AI-powered testing.
QCSD Verification phase swarm for CI/CD pipeline quality gates using regression analysis, flaky test detection, quality gate enforcement, and deployment readiness assessment. Consumes Development outputs (SHIP/CONDITIONAL/HOLD decisions, quality metrics) and produces signals for Production monitoring.
QCSD Development phase swarm for in-sprint code quality assurance using TDD adherence, code complexity analysis, coverage gap detection, and defect prediction. Consumes Refinement outputs (BDD scenarios, SFDIPOT priorities) and produces signals for Verification.
QCSD Production Telemetry phase swarm for post-release production health assessment using DORA metrics, root cause analysis, defect prediction, and cross-phase feedback loops. Consumes CI/CD outputs (RELEASE/REMEDIATE/BLOCK decisions, release readiness metrics) and produces feedback signals to Ideation and Refinement.
