Training for Snowflake Cortex Masterclass Hands-On

Explore Snowflake Cortex components—Snowpark ML, ML-powered classes and functions, and hosted LLMs—while introducing data science basics with Pandas, exploratory data analysis, data profiling, visualizations, and scikit-learn transformers.

Explore quick correlation heatmaps for numeric features in a housing CSV using Python, Streamlit, Matplotlib, and Seaborn to render a symmetrical matrix with selected columns.

Discover the basics of machine learning, including features and labels, training a model, and distinguishing supervised, unsupervised, and reinforcement learning, with notes on regression, classification, clustering, and deep learning.

Navigate the end-to-end machine learning pipeline from data collection and exploration to wrangling, preprocessing, training, cross-validation, and serving, with drift monitoring in Snowflake.

Explore ml pipeline architectures from in-notebook data exploration with pandas to server-side training and deployment in Snowpark and Cortex, including transformers, estimators, and UDFs.

Review the foundation of Snowpark ML within the Snowflake Cortex Masterclass, contrasting it with ML functions and LLM topics, and note the historical use of scikit-learn on Snowflake datasets.

Generate synthetic time series data illustrating the trend, seasonality, and noise, then consider autocorrelation, impulses, exogenous variables, and non-stationary patterns.

learn how regression and classification define supervised machine learning, and how to generate synthetic data with sklearn make_regression and make_classification, creating numeric and categorical features and a binary target.

Generate realistic synthetic data with Faker by creating 10,000 records featuring name, country, city, state, and a 100–10,000 value, assembled into a pandas data frame.

Learn practical data access: download and decompress public datasets from S3 or GitHub, load with pandas read_csv, handle local and remote files, and prepare features and targets for modeling.

Explore simple Jupyter notebooks that demonstrate dataset splitting by column and by row, separating X features from the Y target and preparing train, test, and optional validation sets.

Master data collection for machine learning by loading CSVs, exploring data lakes, generating synthetic data, and leveraging public datasets from Kaggle, Wikipedia, IMDb, and scikit‑learn's iris.

Generate 1000 rows of fake but realistic data by using faker to create city and state pairs, after installing faker with pip and running a simple code cell.

Explore data exploration as the phase between data collection and preprocessing, using data profiling and visualization to identify patterns, outliers, and correlations.

Apply data exploration with a pearson correlation heatmap in a pandas data frame, masked to the upper triangle, visualized with seaborn to interpret dependencies among housing features.

Explore pandas data profiling with describe, data types, and non-null counts, then generate a Titanic profiling report with distributions, correlations, and an HTML export.

Explore how Pandas Profiling generates data visualizations to help you quickly understand datasets, columns, and values, guiding decisions on missing values and encoding.

Explore the data wrangling process from discovery to publishing, focusing on pre-processing, cleaning, imputation, normalization, and feature engineering to prepare data for model training.

Learn to load local data with pandas, perform in-memory feature engineering—selecting, dropping, filtering, and aggregating—using DataFrame operations, and compare with Snowpark data frames.

Preprocess and transform data for machine learning by cleaning missing values with imputation, encoding categorical features with one-hot encoding, and normalizing numeric columns.

Demonstrate data preprocessing with a pipeline and column transformer, handling numeric and categorical features with simpleimputer, standard scaler, and one-hot encoding, renaming columns to upper case and applying 10% sample.

Explore basic machine learning on datasets with Snowpark ML and Snowpark Machine Learning, including reimplemented scikit-learn, xgboost, and LightGBM for performance. Learn how pushdown propagates Python code to virtual machines.

Summarize text quickly using Snowflake Cortex ML Functions with the snowflake.cortex.summarize call. Wrap multi-line inputs in double dollar signs and run the function to generate a concise summary.

Learn how to train models in an ml pipeline with data preprocessing, transformers and estimators, data segregation, cross-validation, and hyperparameter optimization to build reliable regressor or classifier models.

Train a regression model with a pipeline that integrates data preprocessing, imputers, encoders, and a random forest regressor to predict the median house value from housing features.

Train a binary classifier on telescope data with a pipeline (imputer, standard scaler, gradient boosting classifier) and evaluate 10% data, noting training vs evaluation accuracy and score versus predict.

Explore manual hyperparameter optimization for a random forest regressor on the California housing dataset by iterating six parameter combinations of n_estimators and max_depth and comparing training scores.

Split data into training and evaluation sets, reserve a validation dataset, and apply cross-validation such as k-fold and leave-one-out to tune and evaluate models without using test data for training.

Explore GridSearchCV for regression by tuning hyperparameters with a parameter grid, evaluating combinations with cross-validation, and identifying the best params and score.

Explore model validation with randomized search cv for classification, comparing it to grid search cv in scikit-learn, and tuning hyperparameters like learning rate and n estimators.

Explore how model validation uses cross-validation and automatic validation set separation to test hyperparameter grids, with GridSearch and GridSearchCV selecting the best score.

Learn model evaluation in regression through hyperparameter optimization, cross-validation, and grid search, using a diamond price dataset to compare performance metrics like mean squared error and r squared.

Demonstrate classification evaluation in Python with a 10,000-sample binary setup, train/test split, and metrics like accuracy, precision, recall, and F1, plus a confusion matrix for binary and multi-class cases.

Save and load trained models with pickle and joblib, test predictions, and explore model serving and deployment, batch and near real-time scoring, plus MLOps and Snowpark ML implications.

Sign up for a free Snowflake trial by visiting signup.snowflake.com, choose enterprise edition, select AWS as provider, complete email verification, and access Snowsight, the Snowflake web UI.

Explore Snowflake Cortex Masterclass and Snowpark Machine Learning, the Snowpark DataFrame pushdown mechanism, for feature engineering on large data using Python UDFs and stored procedures.

Use the Snowflake web UI (Snowsight) to upload files by selecting database schema and creating a table from file, with the wizard inferring names, types, and a temporary file format.

Explore Snowpark components, including the data frame API and Python-focused tooling, learn about UDFs, stored procedures, UDTFs, and Snowpark ML, plus container services and lazy push-down.

Define and call Snowflake Cortex Masterclass Hands-On stored procedures and Python-based user-defined functions with Snowpark, exploring UDFs, UDTFs, session handling, and a VS Code workflow.

Discover how Snowpark for Python moves business logic to the Snowflake server, enabling Python code via a dataframe API to run as stored procedures or UDFs.

Explore defining and registering Snowflake stored procedures and functions from Python via Snowpark, including anonymous and named options, and translate Python code into SQL for UDFs and UDTFs.

Learn how vectorized user defined functions optimize performance in Snowflake cortex masterclass hands-on by batching inputs with pandas data frames, comparing vectorized and regular UDFs on server side processing.

Select a Python runtime version and pin specific package versions in requirements.txt to ensure Snowflake deployments match locally installed libraries for reliable store procedures.

Explore how Snowpark powers ML pipelines by using Snowpark dataframes for wrangling, training via server-side stored procedures, serving with vectorized UDFs and batching, and choosing standard or optimized warehouses.

Populate a Snowflake data warehouse by creating a test database and schema, then populate time series tables with insert statements and create or replace table as select using json.

Explore synthetic data generation in Snowflake using table functions and a template to generate time series data for regression and classification, including random, uniform, and sequence distributions.

Generate realistic synthetic data in Snowflake using the faker library in a Python worksheet, creating a Snowpark dataframe and saving 10,000 fake rows to a table.

Learn how to connect to Snowflake from Python using the SnowflakeLoginOptions and Session classes from Snowflake ML, via a SnowSQL config dictionary with accountadmin, compute_wh, xsmall, and test public schema.

Learn multiple methods to upload a csv file into a Snowflake table, using the wizard and a no SQL command line interface, with stages, file formats, and copy into.

Master uploading data to Snowflake with Python and Snowpark by creating a session, executing put and copy into, and viewing results in a pandas dataframe and Streamlit app.

Upload and query an external stage file in Snowflake using an S3 public dataset; create a CSV file format, infer schema, and load diamonds data into a table.

Upload multiple datasets and manage stages and tables for sentiment analysis with IMDb reviews. Handle gzip compression, decompress to required formats, and validate training vs test data for model readiness.

Explore extracting a 10% sample from Snowflake data via Snowpark and SQL sampling, compare in-memory pandas sampling with SQL-based sampling, and use Bernoulli, system, or block methods for repeatable results.

Split data into training and test subsets with Snowpark data frame's random_split, using a 90/10 split and optional seed, for classification workflows with train/test views.

Explore ingesting data in Snowflake by uploading files into tables, splitting data into rows or columns, achieving data segregation, and generating synthetic or sampled data for data engineers and scientists.

Learn to generate a correlation heatmap directly in Snowflake by building a Streamlit app, using Matplotlib and Seaborn to visualize the housing table quickly and reliably.

Explore a parquet weather dataset in Snowflake using Snowsight charts and dashboards, infer data types, load and query a weather table, and visualize time-series of temperature and wind.

Explore Snowflake partner notebooks in Deep Note, connect to Snowflake via premium integration, run SQL and Python, and create visualizations while noting CSV-based free usage and a 14-day trial.

Explore Snowflake notebooks in private and public previews, writing and executing SQL or Python at the cell level in SnowSight, with Snowpark ML, Streamlit visuals, and native SQL queries.

Explore data in snow site with built-in charts and dashboards and per-column data profiling. Use Streamlit integrations to build and share maps and visualizations with stakeholders.

Learn fast data profiling in Snowflake by using an SQL worksheet to select all data from your table or query results, instantly generating column-level statistics as a data profile.

Compare pandas data frames with Snowpark data frames, noting that pandas loads all data into memory. Snowpark avoids this and aligns with PySpark as Snowflake's inspired evolution.

Compare pandas and snowpark data frames in snowflake: pandas loads data into memory and executes operations immediately, while snowpark builds a lazy SQL plan and runs only when shown.

Learn feature engineering with pandas data frames via the python connector for Snowflake, loading data into memory, creating calculated columns, filtering with in, and performing mean aggregations.

Explore feature engineering with Snowpark dataframes by transforming a housing table, adding calculated columns, and filtering data. Learn server-side execution with create table as select and memory-efficient transformations.

Compare pandas vs Snowpark data frames on a large Snowflake line item table to test scalability and in-memory loading differences using Python worksheets.

Explore how Snowpark data frames build lazy transformations and generate SQL queries for execution on Snowflake, and compare their behavior with pandas and PySpark.

Python worksheets save you time and money, but they are limited because you cannot pass parameters to functions right away. An article explains how to do it; leave a review.

Explore how a simple Python sequence on a Snowpark DataFrame yields a single SQL statement, a create table as select with two columns, revealed by the query history.

Explore data preprocessing with snowpark data frames in the Snowflake cortex masterclass hands-on, loading CSV from external stages, renaming columns to uppercase, normalizing values, and noting pandas-based transformers in use.

Train a sentiment analysis model on IMDb reviews in local mode with Snowpark and Snowflake ML, converting sentiment to binary and using a count vectorizer with a linear SVM classifier.

Explore training a sentiment analysis model by deploying a Python function as a store procedure in Snowflake, executing on the server side with Snowpark, using CountVectorizer and an SVC classifier.

Explore end-to-end ml pipelines with Snowpark and Cortex, from data sourcing and wrangling to model training, serialization, deployment, and real-time serving using stored procedures and vectorized udfs.

Split the sentiment analysis training function into a separate module, upload it to a file stage, and register a store procedure with imports for remote mode.

Train a regression model for California housing values using a scikit-learn pipeline with imputation, one-hot encoding, and a random forest regressor, deployed as a stored procedure in Snowpark.

Explore model serving with a predict function that loads a name stage model and supports batch and near real-time predictions, using vectorized Python UDFs for concurrency.

Explore sentiment analysis of IMDb reviews, train and test datasets, and deploy a snowflake model using near real-time and batch vectorized UDFs, with joblib loading and caching.

Combine SQL and Python with Snowpark data frames to serve sentiment predictions via UDFs and vectorized UDFs. Compare sentiment flags to labels to assess accuracy.

Load and serve a trained regression model for California house value predictions using a vectorized udf in Snowflake, caching the model, handling test data, and evaluating predictions.

Leverage cache tools to cache function results in memory using a decorator, speeding model loading by memoizing udf calls and reducing disk access with eviction strategies.

Compare regular and vectorized UDFs in Snowflake using Snowpark dataframes, compute mean and standard deviation of account_balance for 15 million customers, and show vectorized UDFs are faster for repeated calls.

Explore Snowflake Cortex and Snowpark ML APIs for modelling, operations, and mlops, data access, with file exports to PyTorch or TensorFlow; include scikit-learn, xgboost, lightgbm wrappers and a model registry.

Snowpark ML provides modeling, operations, and data access APIs wrapping scikit-learn, xgboost, and lightgbm, translating transformers to SQL for distributed execution. Includes automatic stored procedures and a model registry.

Explore data access with the sf file system in snowpark ml, from local files to stage data, and prepare a pandas data frame for PyTorch and TensorFlow.

Create file sets from Snowpark data frames or queries, snapshot them as parquet files in a server-side encrypted stage, and feed the data to TensorFlow and PyTorch via connectors.

Learn to load excel data into snowflake using openpyxl, pandas, and snowpark data frames, from local files and stage files, and to push it via a stored procedure in sql.

Explore snowflake ml data access via file system and file sets, with framework connectors to expose data as TensorFlow records or PyTorch tensors from a server-side encrypted stage.

Explore distributed preprocessing with Snowpark ML versus scikit-learn, comparing in memory pandas workflows to server-side SQL transformations, including ordinal encoding and min-max scaling on large datasets.

Encapsulate scikit-learn preprocessing as a Snowflake stored procedure to compare Snowpark and Snowpark ML for distributed preprocessing with ordinal encoder and minmaxscaler on the regression data set.

Explore distributed preprocessing with Snowpark ML through notebooks that apply min-max scaling, ordinal and one-hot encoding to diamonds data, build pipelines, save models, and analyze correlations.

Snowpark ML distributes pre-processing transformers by translating fit and transform into SQL executed in the Snowflake engine, enabling parallel processing for one hot encoder and label encoder.

Learn the fastest, easiest way to deploy any Python code to Snowflake's server side via Snowpark DataFrames and a Python worksheet in Snowsight, running in the virtual warehouse.

Snowpark ML now provides wrapper classes for scikit-learn and GBM libraries, simplifying model training by auto-managing store procedures and stages, so you call classifiers or regressors from outside like scikit-learn.

Compare Snowpark and Snowpark ML for model training on Snowflake, using Python worksheets, scikit-learn and XGBoost, with ad hoc stored procedures and on-server, lazy evaluation to minimize data transfer.

Executes an end-to-end machine learning workflow with snowflake cortex and snowpark ml, using gradient boosting, imputer, and standard scaler on telescope data for a classification model with train-test accuracy.

Explore model training with Snowpark ml by contrasting estimators and transformers; transformers execute as SQL, while calling fit creates python stored procedure inside the estimator, executed in a secure sandbox.

Explore the estimator pattern in Snowpark ML by fitting an XGBClassifier from a Snowpark DataFrame, using input, label, and output column names, then calling fit on the training data frame.

Explore the roadmap to Snowpark ML and compare prior approaches with the current Snowpark ML and Cortex, highlighting ML pipelines and datasets along the journey.

Explore distributed hyperparameter optimization using grid search and random search with scikit learn and Snowpark ML, comparing in-memory vs store-proc execution on Snowflake for California housing data.

Compare distributed hyperparameter optimization with Snowpark ML versus the old Snowpark approach using scikit-learn in Snowflake, via grid search and random forest, noting speed and cost.

Explore distributed hyperparameter optimization with grid search CV using Snowpark ML and XgbClassifier, a gradient-boosted classifier, on a 1-million-sample in-memory dataset, achieving a best score around 92%.

Explore distributed hyperparameter optimization in snowpark ml, using grid search and random search with cross validation across multi-node warehouses, parallel model training, and evaluation metrics.

Explore distributed metrics in Snowpark ML and sklearn, highlighting classification metrics like precision, recall, and F1, and show how Snowflake runs SQL-based, server-side evaluation on large datasets.

Demonstrate fitting an xgbclassifier with scikit-learn in a Python workflow and compare snowpark versus snowpark ml for distributed metrics on training and test data, including accuracy, precision, recall, and F1.

Explore distributed metrics with Snowpark ML through a notebook experiment using a transformed, one-hot encoded dataset for regression, applying Gridsearchcv with distributed hyperparameter optimization and evaluating with MAPE.

Explore distributed metrics for model evaluation in Snowpark ML, including correlation, covariance, accuracy, f1, recall, precision, log loss, mape, mse, for classification and regression tasks, and parallel hyperparameter optimization.

Explore end-to-end ML operations with Snowpark ML, including model registry, deployment and serving, while monitoring drift, retraining, and versioning across batch and real-time workflows.

Train a random forest regressor on preprocessed diamonds data, then log the trained model version to Snowflake's model registry using log_model, enabling versioned deployment.

Explore snowflake cortex model registry in depth, logging and versioning models, managing metadata and tags, and querying models and versions via show commands and registry API.

Serve a saved model from the registry by passing a test data frame to its predict function, optionally selecting a version, and compare predicted prices to actual.

Explore Snowflake Cortex model registry across providers and model types, log and serve models, and predict with scikit-learn, PyTorch, and TensorFlow.

Identify prediction functions for a specific version of a classification model saved in the model registry with Snowpark ML by using the Models view or the Models folder.

Explore how Snowpark ML handles training and inference with Snowpark optimized warehouses, the model registry, and cost considerations across storage and compute.

Set the warehouse auto suspend to 1 minute to save money. Programmatically create WH1 with initial state suspended and minimum size; the wizard defaults to 5 minutes auto suspend.

Set auto suspend to one minute in Snowflake warehouses to save money; this tip applies to paid Snowflake accounts used with Cortex, not specifically for ML.

Learn how to implement and encapsulate powerful machine learning algorithms as classes and functions in Snowflake, including forecasting, anomaly detection, classification, and feature importance with TOP_INSIGHTS.

Explore the Snowflake classification wizard to perform end-to-end classification, generating SQL, training data selection, evaluation metrics, and ready-to-save predictions with built-in code output.

Explore Snowflake cortex masterclass hands-on and its ML powered functions—classification, time series forecasting, anomaly detection, and contribution explorer—exposed as SQL classes in Snowflake ML.

Explore Snowflake ml classes, including anomaly detection, classification, and forecast, by inspecting built-in classes with show commands, examining functions and procedures, and understanding versioned, private or public preview status.

Learn to work with Snowflake SQL classes: view built-in classes, create and manage objects, call class methods, and handle parameters like search path, and use UDFs, UDDTs, and stored procedures.

Explore how snowflake sql classes create named instances and a singleton account route budget. Activate budgets, set spending limits in credits, and configure email notifications and custom budgets with resources.

Explore how Snowflake Cortex evolves with ML-powered functions, now branded as ML functions, and follow the fast, ongoing updates.

train a binary classifier with snowflake ml using labeled training data and an unlabeled inference set, then evaluate with precision, recall, f1, confusion matrices, and feature importance.

Explore binary and multiclass classifiers in Snowflake, training models and generating predictions from input data. Assess performance with evaluation metrics, confusion matrices, probabilities, and feature importance like interest and rating.

Train a bank classifier on marketing data, splitting into train and inference sets, then generate, save top predictions with false/true probabilities, and review feature importance and the confusion matrix.

Learn classification in Snowflake ml: train a gradient boosting model on a training dataset, predict with an inference data set, and evaluate via confusion matrices, heat maps, and feature importance.

Turn a confusion matrix into a heatmap in Snowsight by switching to chart, renaming the chart to heatgrid, and selecting actual_class and predicted_class.

Forecasting and anomaly detection in snowflake cortex rely on time-series data with timestamps; prepare data by handling gaps, modeling trends and seasonality, and using gradient boosting with 95% prediction interval.

Snowflake Cortex masterclass hands-on teaches preparing sales data for forecasting with timestamps, creating train and test views for multi-series time series, and splitting historical and prediction data.

Train forecasting models with Snowflake Cortex, using time series data, exogenous variables, and multi-series forecasts to predict sales and generate prediction intervals.

Train a forecast model on weather time-series data by converting dates to timestamps, filtering to Albany, and removing duplicates, then forecast temperatures for the next 30 days with 90% interval.

Train a time series forecasting model from historical data with Snowflake Cortex, then forecast with optional exogenous variables and inspect feature importance and evaluation metrics.

Detect anomalies in time series with snowflake cortex’s ml-powered forecasting, using labeled or unlabeled training data to mark test points outside forecast intervals as outliers.

Detect anomalies in sales time series by identifying outliers in jacket and umbrella data, using labeled and unlabeled models with exogenous features like temperature, humidity, and holidays.

Detect temperature outliers in a time series by building a snowflake ml anomaly detection model on an unlabeled data set with a train/test split, handling timestamp format, and scalable warehouses.

Forecasting and anomaly detection share an algorithm, enabling one model to provide both predictions and outlier labels. Outliers lie outside the prediction interval, and training cannot label points as outliers.

Learn how gradient boosting machines form ensembles of weak learners, typically decision trees, to tackle classification (binary and multiclass) and regression tasks for forecasting and anomaly detection within snowflake cortex.

Explore gradient boosting classifier and regressor using scikit-learn with synthetic data, 20 features, 10-fold cross-validation, and evaluate with accuracy and negative mean absolute error.

Explore how the contribution explorer identifies factors that influence a metric by comparing control and test collections. Learn how top insights reveal contributor segments across continuous and categorical dimensions.

Explore the contribution explorer workflow by comparing control and test data across country and department dimensions, using transactions as the metric, and visualizing changes by timestamp to reveal sales drivers.

Analyze loan subscription with binary classification on a marketing dataset, using a top insights function to compare control and test groups and reveal key drivers like age and marital status.

Use Snowflake's top insights to analyze Titanic dataset, converting survival to boolean and evaluating pclass, sex, age, fare, and siblings or spouses and parents or children to reveal survival factors.

Discover why TOP_INSIGHTS is not a time series function, unlike prediction, forecasting, or anomaly detection; the Contribution Explorer has no timestamp and doesn't validate time series integrity in Snowflake Cortex.

Explore Snowflake security, focusing on discretionary and role-based access control for users and roles, including grants, role hierarchy, and primary and active roles.

Define and enforce access rights for classification models in Snowflake, including grant usage and select, establish analyst and consumer roles, and govern model creation, training, and prediction.

Explore forecasting and anomaly detection in Snowflake ML. Learn how access rights and roles—analyst, owner, and consumer—control model creation, training, and execution.

Create and automatically train an ML model for classification, forecasting, or anomaly detection, then grant access to its functions while you remain sole owner and controller, without dropping or retraining.

Explore cost considerations for ML functions in Snowflake, selecting training versus inference warehouses, including Snowpark optimize for training and multi-node for inference to balance compute and concurrency.

Explore Snowflake's built-in LLMs, compare with ChatGPT, and build Streamlit apps using complete, sentiment, translate; review access, cost, Copilot, Universal Search, LangChain, and LlamaIndex.

Pass text to Snowflake's sentiment llm function, run quick queries in seconds, and interpret mood results like 77% positive or 49% bad feeling.

Explore how Snowflake Cortex provides hosted large language model functions for natural language processing, including complete, extract, sentiment, summarize, and translate, with model selection and secure serverless execution.

Explore a quick demo of Snowflake cortex's five language model functions and hosted models, showcasing translation, sentiment analysis, and structured results, while highlighting secure, serverless data governance.

Trace the milestones of data science from statistics and data mining to deep learning and transformers, showing how OLAP and data warehousing led to NLP and generative AI.

Learn how deep learning neural networks underpin generative AI, from artificial and deep neural networks to feedforward networks, CNNs, and RNNs, with LSTM memory for sequences.

Explore generative AI foundations, from NLP and tokenization to embeddings, self-attention transformers, and vector stores, and review applications like text-to-text, text-to-image, and text-to-music.

Examine why deep learning remains underexplored in Snowflake, noting limited coverage and the rise of llms, in a quick checkpoint on deep learning in Snowflake.

Explore local chat applications by integrating OpenAI ChatGPT with Snowflake Cortex, create API keys, run Streamlit translators, and compare Snowflake LM functions with remote GPT models for secure, in-house analytics.

Integrate ChatGPT inside Snowflake Cortex by configuring external access, network rules, and secrets to call the OpenAI API via post requests, using prompt engineering.

Explore two approaches to integrating Snowflake with ChatGPT via the OpenAI API, highlighting safer hosted solutions and the setup of external access integrations, secrets, and Snowpark functions.

Learn to use complete LLM functions in Snowflake Cortex, using simple prompts or message arrays to hosted and self-hosted models, with specialized tasks like sentiment, summarize, and translate.

Explore how the extract_answer LLM function in Snowflake retrieves exact text segments from unstructured data, returning a JSON object with an answer and a confidence score.

Apply the sentiment LM function to text, returning a -1 to 1 sentiment score to reveal mood, with IMDb reviews and prompt-engineered analyses demonstrated under quotas.

Learn how the summarize function converts text into a concise version, with examples from customer reviews and IMDb data, and how prefixes and syntax shape the output.

Translate English to French using the translate LLM function, and explore sentiment analysis and summaries of IMDb reviews with automatic language detection across a dozen languages.

Prioritize specialized LLM functions and avoid the expensive COMPLETE function or Mistral Large; choose the more specialized models for other tasks. Keep parameters consistent and share your review or comment.

Explore building Streamlit apps inside Snowflake using Cortex LLM functions, replacing prompts with translate and complete, executing SQL via Snowpark, and implementing a QnA interface.

Explore how the Snowflake cortex masterclass hands-on shows how the cortex user role grants full access to all LLM functions, and the account admin assigns roles with no granular access.

Explore Snowflake's token-based pricing for LM functions and UI extensions, including per-million-token costs, quotas, context windows, throttling, and warehouse credits.

Compare Mistral Large and Mixtral 8 on Snowflake, noting higher costs per million tokens and a 20x price gap despite similar results (81% vs 71%).

Compare the pricing of Mistral Large with other models to highlight a large discrepancy, and note the introduction of Snowflake Arctic and expected changes.

Explore lm ui extensions in snowsight, including universal search, snowflake copilot, and the document api, to query private metadata and content with natural language and generate sql from text.

Learn how universal search, a natural language, ML-powered tool, quickly discovers and accesses metadata across databases, schemas, tables, views, functions, data marketplace items, and Snowflake documentation.

Explore Snowflake Copilot with a quick demo: create a SQL worksheet, select the TPC-H SF1 schema, and generate, explain, and optimize queries alongside ChatGPT.

Snowflake Copilot, an LM-powered assistant, generates, tunes, and explains SQL from natural language prompts while only accessing metadata, not data, with English support and a helpful right sidebar.

Explore how Snowflake Copilot-inspired workflows generate SQL queries from natural language against Snowflake sample data using LangChain and ChatGPT, including metadata extraction, local Streamlit app, and executing queries.

Explore Snowflake Copilot as a potentially exciting Cortex feature, while noting it can misgenerate SQL statements and users should test carefully and share feedback.

Explore how Document AI with Documenta I lets you open pdfs or documents, ask natural language questions, retain context, and securely extract private data with a pre-trained model.

Learn how llama index builds a private knowledge base from local PDFs and uses a vector store with retrieval augmented generation to answer questions about your data with ChatGPT.

Explore side-by-side comparisons of Snowflake integrations with ChatGPT and the use cases that Snowflake Cortex brings to another level.

Discover added material on SnowPro certifications, including the core prerequisite and advanced track, with a quick intro to the SnowPro Advanced Data Scientist and a practice test.

Understand SnowPro core and five advanced Snowflake certifications, including Architect, Data Engineer, Data Analyst, Data Scientist, and Administrator. Learn exam formats, pricing, renewal, and credentialing via Pearson, Credly, and Accredible.

Discover the SnowPro data scientist certification (DSA-C02), 65 questions in two hours, $375, and two-year validity, covering data science concepts, data pipelining, data preparation, and feature engineering, plus model deployment.

Explore practice questions on fraud detection metrics for classification, dropna in Snowpark data frame, and overloading external functions in Snowflake.

Clone the Snowflake Cortex repository and open it in VSCode; create a Python 3.9 venv, install requirements.txt dependencies like pandas and snowflake-ml-python, and configure a Snowflake trial and ChatGPT account.

Learn to create a free Snowflake trial account, choose enterprise edition, connect via VSCode and SnowSQL, and configure databases, warehouses, and account admin credentials.

Set up and secure your OpenAI account and API key, configure OPENAI_API_KEY in your environment or streamlit secrets, and connect OpenAI to Snowflake for integrated Python and SQL workflows.

Celebrate completing the Snowflake Cortex Masterclass Hands-On and anticipate ongoing updates with new lectures and email announcements from the instructor. Leave a review to help future students discover the course.