Machine Learning Engineers: the all-in-one AI plugin

Skills for machine learning engineers 17

• agency senior ml engineerSets up MLOps pipelines and RAG systems, for example wiring drift alerts and a rollback path for a recommender model.
• data preprocessing advisorGuides cleaning, encoding, scaling, and imbalance handling, for example fixing skewed timestamps before training a demand model.
• machine learning engineerTurns a trained model into fast, cheap inference, for example quantizing a vision model to hit 50ms latency.
• data cleaning guidance assistantFinds duplicates, missing values, and format drift, for example flagging 4,000 duplicate customer rows before analysis.
• debugReproduces and fixes code bugs with root cause analysis, for example tracing a shape mismatch in a training loop.
• ml algorithm selection assistantCompares algorithms and metrics for your data, for example choosing gradient boosting over a neural net on 20k rows.

Plus a start-here assistant, the prompt library and the app builder, and for members the daily brief.

Ready-made prompts 26

• Data Preprocessing & Cleaning3 prompts, like Write a Pandas Data Cleaning Script
• Feature Engineering & Splits5 prompts, like Feature Engineering Ideas
• Model Architecture & Design3 prompts, like Compare AI Model Architectures
• Training & Debugging Runs3 prompts, like Debug A Failing Training Run
• Evaluation & Error Analysis3 prompts, like Choose Metrics to Train and Evaluate Models
• Deployment & Serving3 prompts, like Write a Model Serving API
• Monitoring, Drift & Optimization3 prompts, like Design a Drift Detection Plan

Apps you can build with AI 8

• Managed model training and deployment workspaceKeeps training runs and releases in one place, so you stop hand-copying model artifacts between tools. Demo included.
• AI model test and monitoring workbenchGives you a repeatable test and drift record before release, cutting failed rollbacks. Demo included.
• Raw-source data preparation and stewardship libraryStores your cleaning steps once, so every project starts from the same trusted raw-data pipeline. Demo included.
• Evidence-backed notebook analysis and reporting workspaceKeeps notebook analysis tied to its evidence, so reviewers can follow your reasoning without a re-run. Demo included.
• Source-linked language model workbenchRuns prompts and context in one owned space, so you spend less time on tool switches and review. Demo included.
• AI inference economics workbenchTracks serving cost against measured quality, so you can cut inference spend without guessing. Demo included.
• Source-linked model workspace and admin consoleKeeps models and admin access in one auditable workspace, reducing rented tools and access sprawl. Demo included.
• Source-linked multi-model text and code workbenchConsolidates generation, reasoning, and coding with linked sources, so outputs stay traceable to inputs. Demo included.

AI agents 12

In the download. An agent runs a whole recurring task for you, step by step, and asks you before anything important. Ask your AI: "Run the Experiment Reproducibility Packaging Agent". Each one also has a guided set-up with sample data on its page.

• Experiment Reproducibility Packaging AgentA tested reproduction bundle, with any remaining failures stated.
• Research Method Comparison AgentA fair, source-linked methods comparison.
• Release Checklist and Rollback AgentReach a production deploy where all checks pass and a tested rollback exists.
• AI Model Risk Register AgentA current register of live models with owners, reviews and risk status
• Address Standardization AgentA clean address file where each address is verified or flagged, with duplicates found
• Model Fairness Audit AgentA documented review of how a model performs across groups, with tested mitigations and tradeoffs
• Baseline Versus Complex Model AgentThe simplest model that does the job, with evidence that added complexity was worth it
• Label Quality Audit AgentFind and fix the label errors that hold model quality down, and prove the fix with a retrained comparison
• Data Leakage Detection AgentRemove leaks from splits and features and produce an honest performance estimate
• Champion-Challenger Rollout AgentMove a replacement model from shadow to full traffic only while the evidence supports each step
• Decision Threshold Tuning AgentSet decision thresholds that minimize real-world cost and stay stable as data shifts
• Inference Cost and Latency Review AgentCut serving cost and tail latency without lowering model quality

Connections (MCP servers) 12

A guide in the download. A connection lets your AI work in an app you already use. You add only the ones you use: ask your AI "Which apps can you connect to?" and follow the steps on each connection's page.

• MLflow MCP serverLets your AI look up MLflow experiments, runs, models, and traces and summarize them for you.
• Jungle Grid MCP serverLets your AI estimate, run and check GPU jobs on Jungle Grid, then fetch the results. Official. No install needed.
• Tech Trend MCP serverYour AI can gather fresh research papers, trending code, and FDA device news, then write you a short newspaper-style briefing.
• Jupyter MCP serverLet your AI open, edit and run Jupyter notebooks for you while you watch. No install needed.
• Jupyter Notebook MCP serverLet your AI write and run code inside your Jupyter Notebook, cell by cell, and read the results.
• Optuna MCP serverLet your AI run optimization experiments and show you the results as charts. Official.
• pcq MCP serverLets your AI run, check and compare machine learning experiments in your project and report back what happened.
• Google Colab MCP serverLet your AI open a Google Colab notebook, add and run code cells, and pick a GPU for you.
• Google AI Search MCP serverLets your AI search the web and pull answers from official docs using Google's AI models.
• ZenML MCP serverLets your AI look up ZenML pipelines, runs, models, and deployments, and start new pipeline runs for you.
• Papers with Code MCP serverLets your AI search research papers and find the code, datasets, and results linked to them.
• arXiv MCP serverLets your AI search arXiv for research papers and pull up their details and full text. No install needed.

Your Grok Bot team 12

• Ml EngineerBuild and maintain production ML systems with PyTorch, TensorFlow, and modern MLOps practices.
• Machine Learning Ops Ml PipelineOrchestrate a multi-agent MLOps pipeline from data ingestion to production serving.
• Machine Learning EngineerDeploy and optimize ML models for production inference at scale.
• Ml Pipeline WorkflowEnd-to-end MLOps pipeline orchestration from data prep to model deployment and monitoring.
• Mlops MlflowTrack ML experiments, manage model registry, and deploy models using MLflow.
• Distributed Training Ray TrainScales PyTorch, TensorFlow, and HuggingFace training from one GPU to thousands of nodes across a cluster.
• Inference Serving VllmDeploys and tunes vLLM servers for high-throughput LLM inference with quantization and monitoring.
• Mlops Weights And BiasesTrack ML experiments, visualize training, and manage model registry with Weights & Biases.
• Remote Gpu TrainerDeploy, monitor, and debug long GPU jobs on rented instances with safe teardown and resumable checkpoints.
• Performance MonitorTracks system metrics, detects anomalies, and optimizes resource usage across multi-agent environments.
• Deployment EngineerDesigns and optimizes CI/CD pipelines for faster, safer deployments with automated rollbacks and monitoring, including GitOps and progressive delivery
• ShapExplains machine learning model predictions using SHAP values and visualizations.