Setting Up MLflow for Experiment Tracking in Production

Setting Up MLflow for Experiment Tracking in Production

Imagine a team of five data scientists running 50 experiments every week. Parameters saved in Jupyter notebook comments, metrics in Google Sheets, artifacts in random Google Drive folders. When after a month you need to reproduce the best model with F1=0.924, nobody remembers the exact hyperparameters: learning_rate 0.001, batch_size 32 or 64? Sound familiar? MLflow is an open-source platform that centralizes this chaos: logs parameters, metrics, artifacts, and models for each run. Our engineers (certified ML specialists with five years of experience) deploy MLflow turnkey with PostgreSQL and S3, turning disparate records into a structured system. We've implemented MLflow in 30+ projects, guaranteeing 99.9% SLA for production infrastructure.

Why MLflow became the MLOps standard

MLflow is backed by a community of 10,000+ commits, supports any ML framework (PyTorch, TensorFlow, Hugging Face, scikit-learn), works with Kubernetes, and has APIs for CI/CD. Alternatives like homegrown solutions require constant maintenance and often break. MLflow is a reliable tool that saves teams up to five hours per week on logging. Unlike Bash or Python scripts, MLflow provides a unified UI, flexible API, and integration with 50+ tools.

How to set up MLflow for production

Backend selection

For production we choose PostgreSQL — it handles dozens of concurrent sessions, supports SQL queries, and scales easily.

Artifact storage

Deployment with Docker

Here's a typical Docker Compose configuration we use in production:

services:
  mlflow:
    image: ghcr.io/mlflow/mlflow:v2.14.0
    ports: ["5000:5000"]
    environment:
      - MLFLOW_S3_ENDPOINT_URL=https://storage.yandexcloud.net
      - AWS_ACCESS_KEY_ID=${YC_ACCESS_KEY}
      - AWS_SECRET_ACCESS_KEY=${YC_SECRET_KEY}
    command: >
      mlflow server
      --backend-store-uri postgresql://mlflow:${DB_PASS}@postgres:5432/mlflow
      --default-artifact-root s3://mlops-bucket/mlflow
      --host 0.0.0.0

How to automate logging with autologging?

Simply call mlflow.autolog() at the beginning of your script. Autologging automatically records parameters, metrics, artifacts, and the model for PyTorch, TensorFlow, scikit-learn, Hugging Face. For fine-tuning, use mlflow.sklearn.autolog(log_models=True). This eliminates manual coding for each experiment.

mlflow.autolog()
mlflow.sklearn.autolog(log_models=True, log_input_examples=True)
mlflow.pytorch.autolog(log_every_n_epoch=1)
mlflow.transformers.autolog()

Implementation process

1. Analysis — we discuss your infrastructure, number of users, model types.
2. Design — we choose backend, storage, security level.
3. Implementation — deploy MLflow, configure autologging, integrate with Git and CI/CD.
4. Testing — conduct load testing, verify 99.9% SLA.
5. Deployment — hand over documentation, access credentials, and train your team.

Case study

A computer vision company with 5 data scientists previously logged experiments in Google Sheets. After implementing MLflow, experiment comparison time dropped from two hours to 10 minutes, and the share of reproducible runs rose to 90%. We confirm results with SLA — the standard that ensures transparency.

Timeframe and scope

Roughly 2 weeks to 2 months depending on complexity. Includes: MLflow installation with PostgreSQL and S3, autologging setup, documentation, team training (remote or on-site), and support during the rollout phase (1 month).

Local run and example experiment

For a quick start:

pip install mlflow
mlflow server --host 0.0.0.0 --port 5000

Example experiment logging:

import mlflow

mlflow.set_tracking_uri("http://mlflow-server:5000")

with mlflow.start_run():
    mlflow.log_param("learning_rate", 0.01)
    mlflow.log_params({"batch_size": 32, "epochs": 10, "optimizer": "adam"})

    for epoch in range(10):
        train_loss = train_one_epoch(model, train_loader)
        val_loss, val_acc = evaluate(model, val_loader)
        mlflow.log_metrics({"train_loss": train_loss, "val_loss": val_loss, "val_acc": val_acc}, step=epoch)

    mlflow.log_metric("test_f1", 0.924)
    mlflow.log_artifact("confusion_matrix.png")
    mlflow.log_dict({"feature_importance": feature_imp}, "artifacts/feature_importance.json")
    mlflow.sklearn.log_model(model, "model", registered_model_name="my-classifier")

MLflow Model Registry: managing model versions in production

MLflow includes a built-in model registry — a tool for managing the model lifecycle from experiment to production. Each registered model goes through stages: Staging → Production → Archived.

Transition a model to Production via API:

client = mlflow.tracking.MlflowClient()
client.transition_model_version_stage(
    name="my-classifier",
    version=3,
    stage="Production"
)

Automatic CI/CD rules for promotion: F1 on hold-out set > 0.92 and p99 latency < 100 ms. If a new version fails, it remains in Staging, production is not affected.

For teams of 5+, the model registry reduces the time to roll out a new version from 2–3 hours (manual artifact transfer) to 10–15 minutes via an automated pipeline. Each model version is stored with full metadata: hyperparameters, test-set metrics, dataset hash, and a link to the experiment run.

Conclusion

MLflow is a proven tool for experiment tracking that pays off within the first few weeks. Contact us for an assessment of your infrastructure — we'll prepare an architecture in 2 days. Get a consultation right now. We guarantee results: 99.9% SLA and transparent documentation.

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.