NannyML
30 employees on LinkedIn · 5 known investors
NannyML provides post-deployment monitoring for machine learning models, estimating performance without ground truth labels and detecting data and concept drift to trigger retraining. The platform helps teams maintain ML model reliability in production across enterprises.
Also known as nannyml · NannyML Cloud
Investors · 5
Also in the syndicate · 4
Company profile
researched Aug 2026NannyML develops software for post-deployment data science: monitoring, analyzing and improving machine learning models running in production. Its core open-source Python library estimates model performance when target labels are delayed or unavailable, detects data drift, and links drift alerts back to changes in model performance. The library is model-agnostic, supports tabular classification and regression use cases, provides interactive visualizations, and is distributed through PyPI, conda-forge and Docker.
The commercial product, NannyML Cloud, is built on top of the open-source library and adds infrastructure and capabilities such as concept shift detection, the PAPE performance-estimation algorithm, Slack and email notifications, customizable dashboards, programmatic data collection, metric storage, scheduling of monitoring runs, and segmentation of datasets into subgroups. Users can define a cost-benefit matrix to tie model performance to monetary or business outcomes, rank alerts by importance, and trigger retraining pipelines via webhooks or automate data ingestion through the NannyML Cloud SDK. Documentation also covers probabilistic model evaluation and an experiments module, and tutorials address tabular, text and image data.
Founding story
NannyML was founded in 2020 in Leuven, Belgium, and is led by CEO Hakim Elakhrass, with the aim of helping decision makers by monitoring the decisions AI systems take, how those decisions change over time and how those changes affect the business.
Business model
NannyML operates an open-core model: a free open-source Python library for post-deployment monitoring alongside a paid commercial platform, NannyML Cloud, distributed through cloud marketplaces with free trials, demos and published pricing tiers.
Revenue comes from the commercial NannyML Cloud product, offered either as a managed application provisioned inside a customer's Azure or AWS subscription or as a software-as-a-service deployment on NannyML's infrastructure, with a published pricing page and free trial.
Traction
The open-source repository has about 2.2k stars, 191 forks and more than 1,270 commits, with the library at version 0.13.1 and NannyML Cloud documentation at version 0.24.3. The website displays logos and testimonials from data scientists and data science leaders using the product.
Latest developments
The company's website carries a notice about an acquisition, framed as joining forces to manage the world's automated decisions, though the counterparty and terms are not stated in the available material. NannyML Cloud documentation is at version 0.24.3 and covers deployment on Azure and AWS, a cloud SDK, custom metrics, reporting, probabilistic model evaluation and an experiments module.
▸Full profile — market position, technology, go-to-market, geography, history
Market position
NannyML positions itself around performance-centric monitoring, contrasting its approach with traditional ML monitoring tools that focus on data drift and generate false alarms; its open-source repository has roughly 2.2k GitHub stars and 191 forks.
Rather than alerting on data drift alone, NannyML estimates model performance without labels and connects drift signals to measured performance impact, which the company presents as a way to reduce alert fatigue; it also offers deployment options in which customer data never leaves the customer's own cloud.
Technology
The company's core contributors developed algorithms for estimating model performance without ground truth: confidence-based performance estimation (CBPE) for classification and direct loss estimation (DLE) for regression, plus Probabilistic Adaptive Performance Estimation (PAPE) in the cloud product, described as giving 10% better estimations than CBPE. Multivariate data drift is detected using PCA-based data reconstruction, where reconstruction error is tracked against a threshold derived from a reference period. Univariate drift is detected with statistical tests including Jensen-Shannon distance and L-Infinity distance, corrected for multiplicity, and the same tests are applied to model output and target distribution drift. A Ranker component prioritizes alerts, and concept drift algorithms use recent ground-truth data to determine whether performance changes stem from a change of concept. The library depends on LightGBM.
Go-to-market
Distribution combines a free open-source library (pip, conda, Docker, GitHub) and a community Slack, content marketing through a technical blog and newsletter, and conference presence, with the commercial product sold via the Azure and AWS marketplaces and self-serve free trials, demo bookings and direct founder calls.
Data scientists, machine learning engineers and data science leaders responsible for models running in production, including teams that require deployments where data does not leave their own cloud environment.
Geography
The company is based in Leuven, Belgium, and distributes its cloud product internationally through the Azure and AWS marketplaces.
History
Founded in 2020, NannyML raised a EUR 1.06 million (about USD 1.2 million) round led by Volta Ventures with Lunar Ventures and Belgian angel investors in October 2020. It released an open-source Python library for post-deployment data science and later launched NannyML Cloud, available on the Azure and AWS marketplaces, adding features such as PAPE performance estimation, concept shift detection, segmentation and an experiments module.
Compiled by commissioned research from 8 cited public sources — announcements, filings, and press listed under research sources below.
Key figures
latest reportedCompany-reported or press-reported figures, each dated to when it was claimed — not independently audited.
Timeline · 3
launches, deals, and filingsThe company's website displays a notice inviting visitors to learn more about NannyML's acquisition, described as joining forces to manage the world's automated decisions; the acquirer and terms are not stated.
NannyML Cloud, built on the open-source library, is offered as a managed application deployed within a customer's Azure or AWS subscription or as a SaaS deployment, and is listed on both cloud marketplaces.
The Leuven, Belgium-based startup announced a funding round of EUR 1.06 million (about USD 1.2 million) led by Volta Ventures, with Lunar Ventures and business angels Stijn Christiaens, Jonathan Cornelissen and Lieven Danneels.
$1.2M source ↗
Dated company events from announcements, filings, and press; legal rows summarize public dockets and regulator releases.
▸Research sources · 8
primary sources listed
- NannyMLnannyml.com · web
8 public sources were cited for this profile; the first-party ones are listed here.
Frequently asked questions
- What does NannyML do?
- NannyML builds open-source and cloud software for monitoring machine learning models after deployment, without ground-truth labels.
- Who are NannyML's investors?
- NannyML's investors include Volta Ventures.