JupyterHub for AI/ML: GPU quotas and MLflow integration

JupyterHub for AI/ML: GPU quotas, MLflow integration, and team environments

Working on ML projects in a team often hits the wall of "it works on my machine but not on the server". Different library versions, no GPU access, unsynchronized datasets — chaos that kills productivity. JupyterHub with Kubernetes solves this at the infrastructure level. We have configured dozens of such environments for teams from 3 to 50 people and know all the pitfalls. One client — a company developing NLP models: 12 engineers used to spend up to 4 hours per week syncing environments. After implementing JupyterHub, that time dropped to zero. Turnkey — from image selection to MLflow integration — in 5 business days. We will assess your project for free, contact us.

Problems we solve

• Environment incompatibility. Each developer uses their own package versions. JupyterHub provides a unified Docker image with fixed dependencies. This eliminates "works on my machine" scenarios.
• GPU shortage. Without quotas, one user can hog all resources, leaving colleagues without compute. ResourceQuota and PriorityClass distribute GPUs fairly, and PriorityClass ensures production tasks are not preempted by research ones.
• Data duplication. Datasets are copied to each machine, wasting space and time on sync. A shared PVC in read-only mode solves this: the latest dataset version is available to all users at /data/shared.

Why JupyterHub over separate servers?

How we set up GPU quotas

For fair GPU distribution, we use ResourceQuota on namespaces and PriorityClass for task prioritization. Example configuration:

apiVersion: v1
kind: ResourceQuota
metadata:
  name: jhub-quota
spec:
  hard:
    requests.nvidia.com/gpu: "8"    # Max 8 GPUs at once
    limits.memory: "512Gi"
    requests.cpu: "64"

PriorityClass for GPU: research tasks have low priority, production inference has high priority. This prevents critical processes from being blocked.

How to integrate JupyterHub with MLflow?

Stack: Kubernetes (EKS/GKE), Helm, Docker, PyTorch 2.2, MLflow 2.11, DVC, Great Expectations.

# Add Helm repository
helm repo add jupyterhub https://hub.jupyter.org/helm-chart/
helm repo update

# config.yaml
cat > config.yaml << 'EOF'
hub:
  config:
    Authenticator:
      admin_users:
        - admin
    GitHubOAuthenticator:
      client_id: "your-github-client-id"
      client_secret: "your-github-client-secret"
      oauth_callback_url: "https://jupyter.company.com/hub/oauth_callback"
      allowed_organizations:
        - your-github-org

singleuser:
  image:
    name: jupyter/datascience-notebook
    tag: "python-3.11"
  profileList:
    - display_name: "CPU Standard (4 CPU, 16GB RAM)"
      description: "For EDA and light training"
      default: true
    - display_name: "GPU Instance (1x A100 40GB)"
      description: "For model training"
      kubespawner_override:
        extra_resource_limits:
          nvidia.com/gpu: "1"
    - display_name: "GPU Large (2x A100 80GB)"
      kubespawner_override:
        extra_resource_limits:
          nvidia.com/gpu: "2"
  storage:
    capacity: 50Gi
    homeMountPath: /home/jovyan

# Shared storage for datasets (read-only for users)
singleuser:
  extraVolumes:
    - name: shared-datasets
      persistentVolumeClaim:
        claimName: shared-datasets-pvc
        readOnly: true
  extraVolumeMounts:
    - name: shared-datasets
      mountPath: /data/shared
      readOnly: true
EOF

helm install jupyterhub jupyterhub/jupyterhub \
  --namespace jhub --create-namespace \
  --values config.yaml

For more details on configuration, see the official Zero to JupyterHub documentation.

Resource profiles detailed in the table:

Custom Docker images for ML

FROM jupyter/datascience-notebook:python-3.11

USER root
RUN apt-get update && apt-get install -y \
    libgomp1 \
    && rm -rf /var/lib/apt/lists/*

USER ${NB_UID}

# ML dependencies
RUN pip install --no-cache-dir \
    torch==2.2.0 torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121 \
    transformers==4.38.0 \
    datasets \
    accelerate \
    peft \
    mlflow==2.11.0 \
    dvc[s3] \
    great_expectations \
    lightgbm xgboost catboost \
    optuna \
    shap \
    wandb

# MLflow tracking server URL
ENV MLFLOW_TRACKING_URI=http://mlflow.internal:5000

# DVC remote config
COPY dvc_config /home/jovyan/.dvc/config

MLflow is automatically available from all notebooks via the environment variable. DVC is configured with corporate remote storage. The shared dataset folder with the latest dataset versions is mounted read-only. Git pre-commit hooks are installed globally for code standardization.

Typical result: An ML team of 10+ people works in a unified environment without "works on my machine" issues, with shared access to GPU resources and centralized experiment tracking. This significantly reduces compute costs through GPU utilization.

Process

1. Analysis — study current infrastructure, GPU requirements, data volume.
2. Design — develop architecture: namespaces, images, quotas.
3. Implementation — deploy JupyterHub via Helm, configure authentication.
4. Integration — connect MLflow, DVC, shared storage.
5. Testing — verify scenarios: training launch, logging, data access.
6. Team training — conduct a workshop, hand over documentation.

What's included

Guarantees and experience

Over 5 years of experience in MLOps, 20+ deployments for teams from 3 to 50 people. We guarantee 99.9% SLA availability and environment reproducibility. Order JupyterHub setup for your project — we will provide demo access to a working environment within 2 days. Get a consultation by contacting us.

MLOps: Infrastructure for Training, Deploying, and Monitoring ML Models

The model is trained, metrics — F1 0.94 on validation. Three months later in production, quality drops by 12%. No one knows when — there is no monitoring. It's impossible to retrain quickly — the training script is in a Jupyter notebook of a data scientist who has already left. Data for retraining is collected manually from three disparate systems. About half of the projects come to us with this pain. We build a turnkey MLOps platform: from experiment tracking to automatic deployment and data drift monitoring. We will assess your infrastructure in 1–2 weeks, and in 4–6 weeks you will get a basic MLOps core running in production. Our team has 10+ years of experience in ML infrastructure, over 50 implementations.

How does MLOps infrastructure benefit your ML projects?

Experiment Tracking and Reproducibility

Without tracking, an ML project turns into chaos: it's unclear which checkpoint is better, which hyperparameters were used, which dataset. Reproducing a result a month later is a quest.

Why is experiment tracking the foundation of reproducibility?

MLflow is an open source standard for tracking. It logs parameters, metrics, artifacts (models, graphs), and code. MLflow Model Registry is a centralized model storage with versioning and lifecycle stages (Staging → Production → Archived). Deployment via MLflow Serving or integration with external systems.

Typical initialization in code:

import mlflow

mlflow.set_experiment("fraud-detection-v2")
with mlflow.start_run():
    mlflow.log_params({"learning_rate": 3e-4, "batch_size": 64, "epochs": 10})
    mlflow.log_metric("val_f1", val_f1, step=epoch)
    mlflow.pytorch.log_model(model, "model")

This is the minimum. In production, we add logging of system metrics (GPU utilization, memory), dataset (hash, version), code (git commit hash). Weights & Biases — richer UI, collaboration features, sweep for hyperparameter optimization. MLflow — for on-premise deployment without external dependencies.

DVC (Data Version Control) — versioning of data and models on top of git. Data is stored in S3/GCS/Azure Blob, only metadata (hashes) in git. dvc repro reproduces the entire pipeline from raw data to metrics.

To ensure reproducibility of training, fix random seeds (torch.manual_seed, numpy.random.seed, random.seed) and record them in experiment metadata. Without this, debugging irregular results is painful. Log the dataset version (DVC hash) and git commit — then any experiment can be reproduced down to the byte.

Pipeline Orchestration: Kubeflow, Airflow, Prefect

A pipeline orchestrator becomes necessary when: A 100-line training script in cron is fine for simple tasks. But as soon as you have a multi-step pipeline (data loading → preprocessing → feature engineering → training → validation → deployment if quality above threshold), you need an orchestrator with retry logic, visualization, and alerts.

Kubeflow — Kubernetes-native orchestrator for ML (see Kubeflow). Each step is a Docker container. Supports parallel steps, conditional branches, artifacts between steps. Integrates with Katib (AutoML), KServe (serving), Feast (feature store).

Apache Airflow — more general DAG orchestrator. Wide ecosystem of operators (S3, Spark, DBT, Kubernetes). Easier to deploy if Airflow already exists in the company.

Prefect / Metaflow — less boilerplate. Prefect 2.x with @flow and @task decorators — quick start for small teams.

Typical training pipeline architecture on Kubeflow:

1. Data ingestion component — fetches data from S3/DB, validates schema via Great Expectations
2. Preprocessing component — transformations, normalization, train/val/test split
3. Training component — training on GPU, logging to MLflow
4. Evaluation component — metric calculation, comparison with baseline in Model Registry
5. Conditional deployment — deploy only if new model is better than current by >2% F1

Each component is a separate Docker image. Pipeline is versioned in git. Scheduled run (retraining once a week on new data) or manual.

Model Registry and Lifecycle Management

Model Registry is not just a checkpoint store. It is a centralized system that knows:

• Which model is currently in production (and with what metrics)
• History of all versions with training parameters
• Metadata: dataset, git commit, validation results
• Lifecycle stage: None → Staging → Production → Archived

MLflow Model Registry — standard. For enterprise — Vertex AI Model Registry (GCP), SageMaker Model Registry (AWS), Azure ML Model Registry.

Model promotion through stages: automatically move model to Staging after successful eval, then manual or automatic (during A/B test) promotion to Production. Rollback — switch to previous Production version in seconds.

Serving: From FastAPI to Triton Inference Server

Simple case. FastAPI + PyTorch/ONNX on one server — 80% of production ML deployments are exactly that. Sufficient for most tasks with load up to 100 req/s.

from fastapi import FastAPI
import onnxruntime as ort

app = FastAPI()
session = ort.InferenceSession("model.onnx", providers=["CUDAExecutionProvider"])

@app.post("/predict")
async def predict(request: PredictRequest):
    inputs = preprocess(request.text)
    outputs = session.run(None, {"input_ids": inputs})
    return {"label": postprocess(outputs)}

Triton Inference Server — production standard for high loads (500+ req/s). Dynamic batching, concurrent model execution, model ensemble. Supports TensorRT, ONNX, PyTorch TorchScript, TensorFlow SavedModel.

KServe — Kubernetes-native ML serving with autoscaling, canary deployments, A/B testing out of the box. Scale-to-zero for inactive models — savings on infrastructure up to 40% annually for a project with 10 models.

Monitoring: Data Drift, Model Drift, Infrastructure Metrics

Monitoring — what is usually done last and regretted first. Three levels.

Infrastructure monitoring. Latency (P50/P95/P99), throughput (req/s), error rate (4xx, 5xx), GPU/CPU utilization. Prometheus + Grafana — standard. Alert when P99 latency > threshold or error rate > 1%.

Data drift monitoring. Distribution of input data changes over time. Detect via PSI (Population Stability Index) for numerical features: PSI > 0.2 — strong drift. Chi-squared test for categorical, Kolmogorov-Smirnov test for continuous. Evidently AI — open source library with ready-made drift tests.

Model drift monitoring. If ground truth is delayed (e.g., we know conversion after a week) — monitor real metrics. If not — surrogate metrics: distribution of prediction scores, proportion of confident predictions.

Alerting. Three levels: INFO (minor drift, log it), WARNING (significant, notify team), CRITICAL (quality dropped below threshold — automatic switch to fallback model).

Why is data drift monitoring important?

Without it, you learn about model degradation only from user complaints or ringing SLA. A drift alert allows you to retrain the model in advance, before errors start causing losses. In one of our projects, PSI monitoring detected drift 2 days after a data source change — this saved the campaign.

Feature Store

Feature Store solves the training-serving skew problem. If preprocessing during training and inference is implemented in two different places — divergence is inevitable.

A Feature Store is needed when:

• Several models use the same features
• Features are computed from streaming data (real-time)
• Large team with different people on feature engineering and model training

Feast — open source Feature Store. Offline store (S3 + Parquet) for training, online store (Redis, DynamoDB) for low-latency inference. Feature definitions as code, materialization job syncs offline → online.

Tecton (commercial), Vertex AI Feature Store (GCP), SageMaker Feature Store (AWS) — managed options with less ops overhead.

CI/CD for ML

ML CI/CD is regular CI/CD plus specific ML steps.

ML-specific checks in CI:

• Reproducibility check: run training with a fixed seed, result must match
• Data validation: Great Expectations or Pandera on schema/distribution checks
• Model performance check: automatic eval on holdout, block merge if degradation > threshold
• Latency regression test: inference must meet SLA

GitOps for deployment. Merge to main → CI triggers training → eval → if passes → automatic deployment to Staging → smoke tests → manual promotion to Production or automatic upon successful canary.

Tools: GitHub Actions / GitLab CI for CI, ArgoCD for GitOps deployment on Kubernetes.

What's Included in MLOps Platform Development

We provide a full cycle of work, documentation, and team training.

Total time from audit to full MLOps platform: 3–5 months. Also possible phased launch: basic level (tracking + serving) in 4–6 weeks.

Cost is calculated individually based on data volume, number of models, and infrastructure requirements. Order an MLOps infrastructure audit — get a roadmap in 1–2 weeks. Contact us for a project assessment — we will send a preliminary estimate within 2 business days.

Note: warranty on architectural solutions — 12 months. We provide integration certificates with major cloud providers (AWS, GCP, Azure). During our work, we have not lost a single client after the first implementation — the experience of 50+ successful MLOps projects speaks for itself. Get a consultation on building an MLOps platform today.