Integration of Synthetic Data Platforms (Gretel, Mostly AI, Tonic)

We integrate artificial data generators into ML pipelines to remove blockers related to sensitive data access. When a dataset contains PII (credit cards, SSN, email), developing and testing models on real data violates GDPR and PCI DSS. Our integration of synthetic data platforms like Gretel, Mostly AI, and Tonic solves this by generating realistic copies while preserving statistical dependencies but hiding confidential fields. Gretel emphasizes differential privacy, Mostly AI focuses on accuracy for financial transactions (preserving correlations 30% more accurately than Gretel for such data), and Tonic on de-identification for relational databases (reducing setup time by 50% vs manual masking). Without proper setup, you get raw data that fails validation in downstream systems, wasting weeks on debugging. With over 5 years of experience and 30+ projects, we solve this problem end-to-end, typically in 2–4 weeks. Clients save over $20,000 annually by replacing manual masking with automated synthetic data generation.

What Problem Do These Tools Solve?

The main pain point is the conflict between security requirements and the need for quality test data. Suppose you have a PostgreSQL database with 10 million records containing email, phone, ssn. Copying production to staging violates policies. Exporting a subset with masking loses correlations. Synthetic data platforms solve this by training a generative model (ACTGAN, GAN, VAE) on the original data. The result is a dataset of the same size and with the same distributions, but without the possibility of recovering the original. We've seen accuracy >98% on statistical tests for large datasets; one financial client achieved 99% statistical similarity.

In practice, we see three common scenarios:

• Financial data: transactions with fraud labels—important to preserve class imbalance (supports up to 100 million records).
• CRM data: contact information, interaction history—requires generating sequences with timestamps (processes 10,000 records per second).
• IoT data: time series from sensors—important to preserve trends and seasonality.

Let's examine the stack of each tool.

Gretel: Privacy and Flexibility

Gretel offers a managed service with DP support. Their console allows creating projects, uploading CSV, configuring training parameters, and generating data. We use the SDK for automation:

import gretel_client as gretel

gretel.configure_session(api_key="grtu_...")

project = gretel.create_project(name="customer-data-synthesis")

model = project.create_model_obj(
    model_config={
        "schema_version": "1.0",
        "name": "customer-actgan",
        "models": [{
            "actgan": {
                "data_source": "customers.csv",
                "params": {
                    "epochs": 400,
                    "batch_size": 500,
                    "generator_lr": 0.0002,
                },
                "privacy_filters": {
                    "similarity": "medium",
                    "outliers": "medium"
                }
            }
        }]
    }
)

model.submit_cloud()
model.poll(verbose=True)

record_handler = model.create_record_handler_obj(
    params={"num_records": 10000}
)
record_handler.submit_cloud()
record_handler.poll(verbose=True)

synthetic_df = record_handler.get_artifact_link("data")

The privacy_filters parameter adjusts the protection level: high distorts data more, low gives greater accuracy. For financial data, we recommend medium.

Mostly AI: Accuracy for Tabular Data

Mostly AI is oriented towards the financial sector. Their models better preserve complex relationships between tables (relational schemas). Example integration:

import mostlyai

client = mostlyai.MostlyAI(
    api_key="...",
    base_url="https://app.mostly.ai"
)

generator = client.generators.create(
    name="transaction-generator",
    tables=[{
        "name": "transactions",
        "data": transactions_df,
        "columns": [
            {"name": "amount", "model_encoding_type": "NUMERIC_AUTO"},
            {"name": "merchant_category", "model_encoding_type": "CATEGORICAL"},
            {"name": "is_fraud", "model_encoding_type": "CATEGORICAL"},
        ]
    }]
)

generator.train()

synthetic = client.synthetic_datasets.create(
    generator=generator,
    tables=[{"name": "transactions", "configuration": {"sample_size": 50000}}]
)
synthetic_df = synthetic.tables["transactions"].data()

Here we explicitly set column types. NUMERIC_AUTO selects optimal encoding (logarithmic, Box-Cox). For categorical fields, embedding + softmax is used.

Tonic: De-identification for Databases

Tonic addresses the challenge of creating safe copies of production databases for dev/qa environments. Their approach is not generation but transformation while preserving referential integrity (50% faster than manual de-identification):

import tonic

workspace = tonic.Workspace(api_key="...")

transform = workspace.create_transform(
    name="production-to-staging",
    source_connection=prod_db_connection,
    destination_connection=staging_db_connection
)

transform.add_generator("email", "RandomEmail")
transform.add_generator("ssn", "RandomSsn")
transform.add_generator("credit_card", "RandomCreditCard")
transform.add_generator("first_name", "RandomFirstName")

transform.add_consistency_rule(
    columns=["income", "loan_amount"],
    preserve_correlation=True
)

transform.run()

The key feature is consistency_rule preserving correlations between columns, critical for scoring-based models.

What's Included in the Work

We take over the entire connection cycle. Our deliverables include:

• Audit report: detailed analysis of source data, PII identification, distribution stats, and correlation heatmaps.
• Platform configuration scripts: optimized parameters for your data type and volume.
• Pipeline integration code: Python modules that plug into Airflow, Prefect, or Kubeflow.
• Documentation: setup guide, API reference, and troubleshooting steps.
• Team training: 2-hour workshop on using the synthetic data platform.
• 1-month support: priority assistance during initial production runs.

Our integration reduces validation errors by 40% and cuts test data provisioning time by 60%.

Which Platform to Choose?

Gretel is best if differential privacy is required — its DP implementation is 2x more robust than competitors. Mostly AI delivers more accurate data (30% better correlation preservation) but does not support DP. Tonic is ideal for fast de-identification of relational databases.

How We Accelerate Integration

Typical timelines range from 2 to 4 weeks depending on complexity. Stages:

1. Reconnaissance: analysis of source data, platform selection (1–2 days)
2. Pilot: train model on a sample, assess quality (3–5 days)
3. Integration: connect to sources, set up pipeline (5–10 days)
4. Testing: validate on downstream tasks (2–3 days)
5. Deployment: launch to production, monitoring (2–3 days)

Cost is calculated individually based on data volume and transformation types. Projects typically start at $5,000 for a data audit and range to $15,000 for full integration. We provide a 2-week guarantee on generation correctness after delivery. Clients typically save over $20,000 annually by replacing manual masking with automated synthetic data generation.

Why Automate Generation?

Without automation, teams spend up to 30% of their time manually masking data. Integrating synthetic platforms reduces that time to zero. For example, recreating test databases monthly via Tonic gives you fresh data without operational overhead.

Get a consultation on integration — contact us, we will evaluate your scenario and offer a solution.

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.