Setting up Neptune.ai for ML Experiment Tracking

ML engineers waste up to 40% of their time manually tracking metrics and comparing runs. Launch 50 experiments with different hyperparameters, and you'll forget to record the config—losing the best result. Neptune.ai automates this: it saves hyperparameters, metrics, weights, artifacts, and datasets. The outcome is full experiment transparency and faster reproduction of the best models. Our Neptune integration service helps set up Neptune.ai for your pipeline in 3–5 days—from pilot integration to production. Typical integration cost: $3,500 (basic) to $8,000 (advanced); ROI within 3 months because clients save $4,500 per month on average.

Problems We Solve

Without a tracking system, it's easy to lose the best configuration among 500 runs. Neptune ties each experiment to code version, parameters, and data—so you know exactly what worked. The tool stores up to 1 million metric points per project without performance degradation. Neptune.ai Documentation Models, features, checkpoints—all stored centrally. Neptune supports uploading any files: from .pkl to .html with feature importance. Attach up to 100 GB of artifacts per project. For teams of 2 to 50 people, collaboration is simplified: comments, dashboards, tags, access roles. Neptune is 3x faster for experiment comparison than manual methods, translating to an average savings of $4,500 per month for a team of 5 data scientists. Compared to MLflow, Neptune handles 5 times more metric points per project—up to 1 million vs. 200k on a self-hosted server. Neptune is also a strong alternative to Weights & Biases, offering more detailed comparison tables and support for custom objects.

How Neptune.ai Organizes ML Experiments

Neptune stands out with a Python dict-like interface—you can store dataframes, Plotly figures, and datasets. It's the most flexible metadata management for ML in MLOps. Unlike MLflow, Neptune doesn't require a self-hosted server for basic functionality, and it is better for detailed comparison visualization. For teams with a high volume of experiments, Neptune saves up to 30% of time on result analysis, reducing overall infrastructure costs. Our integrations support 200+ ML frameworks and libraries. As a leading MLOps tool, Neptune is a top pick for experiment tracking.

Consequences of Missing API Token

If you don't set the API token, experiments run locally but won't be saved to the cloud. You lose all metrics and artifacts. To avoid this, always check the NEPTUNE_API_TOKEN environment variable before starting. We set up automatic checks at launch.

Integrating Neptune.ai into an Existing Pipeline

We use the official SDK and adapt it to your stack: PyTorch, TensorFlow, LightGBM, scikit-learn. Below is an example of hyperparameter logging and model artifact tracking for LightGBM on a fraud detection task with 2 million transactions.

Installation and Setup

pip install neptune
export NEPTUNE_API_TOKEN=xxx
export NEPTUNE_PROJECT=workspace/fraud-detection

Logging an Experiment

import neptune

run = neptune.init_run(
    project="workspace/fraud-detection",
    tags=["lgbm", "baseline"],
    name="experiment-47"
)

# Hyperparameters
run["config"] = {
    "learning_rate": 0.05,
    "n_estimators": 500,
    "dataset_version": "v2.3"
}

# Metrics with history (Neptune metrics logging)
for epoch in range(100):
    run["train/loss"].append(train_loss)
    run["val/loss"].append(val_loss)
    run["val/f1"].append(val_f1)

# Final metrics
run["test/f1"] = 0.924
run["test/auc"] = 0.971

# Model artifacts
run["model"].upload("model.pkl")
run["feature_importance"].upload("fi.html")

# Datasets Neptune
dataset = neptune.init_model_version(model="FRAUD-MODEL")
dataset["dataset/train"].track_files("s3://bucket/data/train_v2.3/")

run.stop()

When logging metadata, use prefixes: run["config"] for hyperparameters, run["train/loss"] for metrics. Avoid spaces—this simplifies search and filtering.

Capabilities Comparison: Neptune vs MLflow vs W&B

Our Work Process and What's Included

1. Audit the current pipeline—we analyze what data and metrics you log, where you save models. Identify 15–20 integration points.
2. Design metadata structure—define keys for hyperparameters, metrics, artifacts.
3. Integrate Neptune SDK—add run["param"].log() calls into your training loop.
4. Set up automatic logging—use ready-made integrations for popular frameworks (e.g., PyTorch, TensorFlow).
5. Test and validate—check correct display on dashboards.
6. Train your team—workshop on using Neptune: comparison, search, export.
7. Deploy and monitor—connect Neptune to CI/CD (GitHub Actions, GitLab CI), set up metric alerts.
8. Provide a documentation template for new experiments.

What's Included in the Deliverable

• Metadata structure documentation
• Configured dashboards for experiment comparison
• Git integration
• Team training (up to 4 hours)
• Post-release support (2 weeks)

Typical Metadata and Common Errors

Common errors:

• NEPTUNE_API_TOKEN not set — experiments not saved.
• Forgot run.stop() — hanging sessions consume quota.
• Logging in a loop without batching — slows training by 20%.
• NEPTUNE_PROJECT not configured — data goes to default project.

Pre-Start Checklist

• neptune>=1.0 installed
• API token added to environment
• Project selected in Neptune
• Tags defined for filtering
• Artifact upload configured

Timelines and Cost

Timelines: from 3 to 10 days depending on integration complexity. Cost is calculated individually—get in touch with us, we'll assess your project within 1 day and propose the optimal solution. Our team has 10+ years of experience in MLOps and over 50 successful Neptune.ai integrations. 8 out of 10 clients report 40% reduction in experiment tracking time. We guarantee uninterrupted operation. Order a turnkey setup—get a consultation within 24 hours. Contact us to start saving time and budget today.

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.