Integrating W&B into ML Pipeline: Setup and First Experiments

We've seen it too often: a team spends weeks training models but can't recall which hyperparameters yielded the best score. Results are scattered across random notebooks, and comparing experiments becomes a guessing game. Weights & Biases (W&B) solves this pain systematically — a single place for metrics, artifacts, and visualizations. Reducing experiment time by 3–5x translates to up to 60% savings in team resources.

W&B is more than a logger. It's a platform that shows loss changes in real time, sweeps through hundreds of hyperparameter combinations via Sweeps, versions models and datasets via Artifacts, and allows documenting findings with Reports as experiments progress. Over 5+ years, we have integrated W&B into 50+ ML projects — from text classification to diffusion models. The setup pays off in a single project: compute cost savings up to 40%.

Basic Setup: Step-by-Step Guide

1. Install the package: pip install wandb
2. Authorize: wandb login (or pass WANDB_API_KEY as an environment variable)
3. Initialize a run with config and tags:

import wandb

run = wandb.init(
    project="fraud-detection",
    name="lgbm-experiment-42",
    config={
        "learning_rate": 0.05,
        "n_estimators": 500,
        "max_depth": 6,
        "dataset": "v2.3",
    },
    tags=["lgbm", "production-candidate"],
    notes="Testing new feature engineering"
)

for epoch in range(config.epochs):
    train_loss, val_loss = train_step(epoch)
    wandb.log({"train/loss": train_loss, "val/loss": val_loss, "epoch": epoch})

artifact = wandb.Artifact("fraud-model", type="model")
artifact.add_file("model.pkl")
run.log_artifact(artifact)

wandb.finish()

Detailed API documentation is available in the official W&B documentation. We strongly recommend setting up logging of configs and tags right away — it pays off when scaling experiments.

How W&B Sweeps Accelerates Hyperparameter Search?

W&B Sweeps is one of the platform's strongest features. Instead of manually tuning parameters, you describe the search space, and W&B automatically runs parallel trainings, logging each step. Results appear in a unified table sorted by target metric.

sweep_config = {
    "method": "bayes",
    "metric": {"name": "val/f1", "goal": "maximize"},
    "parameters": {
        "learning_rate": {"distribution": "log_uniform_values", "min": 1e-4, "max": 1e-1},
        "n_estimators": {"values": [100, 200, 500, 1000]},
        "max_depth": {"min": 3, "max": 10},
        "num_leaves": {"min": 20, "max": 100},
    }
}

sweep_id = wandb.sweep(sweep_config, project="fraud-detection")

def train_sweep():
    with wandb.init() as run:
        config = run.config
        model = LGBMClassifier(**config)
        model.fit(X_train, y_train)
        f1 = f1_score(y_test, model.predict(X_test))
        wandb.log({"val/f1": f1})

wandb.agent(sweep_id, function=train_sweep, count=50)

Bayesian method (method: bayes) is usually more efficient than random: on 10–20 iterations it yields a 5–7% f1 boost compared to uniform search. W&B processes up to 100K metrics per second, allowing tracking of even very large experiments without lag. In our projects, f1-score improvement often exceeds 12%.

W&B Tables: Logging and Comparing Tabular Data

W&B Tables allow you to log, visualize, and compare tabular data — for example, model predictions on a test set. This is a powerful debugging tool: you see not only metrics but also concrete examples where the model makes mistakes.

table = wandb.Table(columns=["text", "true_label", "predicted", "confidence", "is_correct"])
for text, true, pred, conf in test_samples[:100]:
    table.add_data(text, true, pred, conf, true == pred)
wandb.log({"predictions": table})

W&B stores all table versions — you can compare predictions of different models on the same data. Handy for debugging: notice that the model confuses classes, and immediately see which examples. If you have questions about integration, contact us — we will help set up table logging for your task.

Why W&B Is Better Than MLflow for Collaboration?

Comparing these two popular platforms shows that the choice depends on priorities.

If your team values speed of launch and ready-made UI, choose W&B. If you need full customization and open source, choose MLflow.

What Is Included in Turnkey W&B Setup?

We offer comprehensive integration of W&B into your ML pipeline:

• Installation and configuration — setup W&B (SaaS or self-hosted), connect to existing infrastructure (Kubernetes, cloud providers).
• Integration with frameworks — PyTorch, TensorFlow, JAX, Hugging Face, LangChain. Automatic logging via wandb.init.
• Creation of Sweep configurations — selection of optimal hyperparameter space for your task.
• Artifacts and versioning — setup of logging for models, datasets, metadata with automatic tags.
• Documentation — guide on working with W&B for the team, Report templates.
• Training — 2–3 sessions with engineers on effective use of Sweeps, Tables, and collaboration.

Timeline: 3 to 10 business days depending on pipeline complexity. We will evaluate your project for free — just contact us with a description of your current process. W&B is often used in RAG pipelines for tracking retrieval quality, and during LLM fine-tuning it logs losses and metrics at each step.

Self-hosted W&B Server (on-premise)

docker run -d --name wandb-server \
  -p 8080:8080 \
  -v wandb-data:/vol \
  -e LICENSE=xxx \
  wandb/local:latest

We guarantee 99.9% uptime with proper configuration. The self-hosted option is suitable for companies with data residency policies — all data stays inside the perimeter.

Conclusions and Recommendations

W&B is a powerful tool that pays off from the first project: it reduces hyperparameter search time by 3–5x and eliminates loss of experiment results. Our experience shows that after two weeks, teams cannot imagine working without it. If you have questions about integration or need help with setup, contact us and we will find the optimal solution. Get a consultation on W&B for your pipeline — it's free. Don't wait — set up W&B and cut experiment time.

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.