Meet the team
mloda is built in Berlin by a team that has spent years untangling data infrastructure - and decided to fix it at the root.
“Striving to make data and algorithmic pipelines robust, reliable, and finally painless.”
Everywhere I've worked - data science, AI, software engineering - the problem was never the models. It was everything before them: broken pipelines, inconsistent features, fragile infrastructure. mloda is my answer to that chaos: a lightweight, no-nonsense framework for rock-solid pipelines from day one.
“Good science runs on trustworthy data. So does good AI. mloda makes that the default.”
As a researcher at HU Berlin and the Max Planck Institute, I worked with data every day - and learned how much good science depends on data you can trace and trust. While Tom architects the framework, I bridge deep tech and the market, turning complex compliance challenges into clear, strategic narratives.
Developer stories
How mloda changes the day-to-day of the people building on it - from laptop prototypes to audited production pipelines.
Monday: Prototype → Friday: Production
Meet Sarah, a data scientist on a mission to predict churn. Watch the same feature code scale from her laptop to production, serving millions.
Sarah suspects that customers who reduce their spending by 40% over 30 days are likely to churn. To validate this quickly, she grabs a 10 k‑row sample on her laptop, defines churn features with mloda and runs them in minutes using Pandas.
The hypothesis shows promise! Early results look promising, but Sarah needs proof and wants to check it with 50M+ records. Without changing her feature definitions, she switches to a Spark cluster. The same churn logic now processes the distributed data at scale.
Model validated! Marketing needs real-time churn scores for their retention campaign. Sarah deploys the exact same features as an API. Now serving 10K requests/second with sub-millisecond latency using PyArrow.
Second-Generation: Built for the Age of AI Assistants
Unlike first-generation tools retrofitted for AI, mloda was architected from day one for AI integration. Through MCP servers and template-based FeatureGroups, AI assistants can create, discover, and execute features as naturally as calling a function.
Ask your assistant to build a churn‑prediction feature and mloda’s stable templates handle the rest. The AI produces a well‑structured FeatureGroup, and mloda’s testing architecture ensures the output is valid and ready for production.
Your FeatureGroup instantly becomes available as an MCP tool. Any AI assistant can now execute feature calculations, discover related features, and chain operations together through standard MCP protocol. This isn't an API wrapper, it's native AI-to-mloda communication.
When AI calls mlodaAPI.run_all(), it's not generating code, it's executing real transformations. AI can orchestrate complex feature pipelines, combine multiple FeatureGroups, and deliver actual results. Each successful execution teaches the AI new patterns, making the next interaction even more powerful.
Community Plugins: Alex's Fraud Detection Journey
Join Alex, a startup founder racing to beat enterprise competitors. Discover how the community's shared plugins let his 3-person team deliver features that rival 50-engineer teams, while keeping their secret sauce private.
Alex needs fraud detection features fast. Instead of building from scratch, he discovers 50+ fraud-related FeatureGroups in mloda's plugin repository. He loads FraudPatternDetector and what would take weeks is running in minutes.
Alex extends community plugins with his company's unique signals using mloda's Extenders. He adds monitoring and validation, then shares the wrapper back (keeping proprietary logic private). His contribution helps 1000+ engineers.
Alex's production system combines 5 community plugins with his custom logic. FraudPatternDetector, VelocityChecker, GeoAnomalyDetector orchestrated by mloda core. He's preventing fraud using collective intelligence from the entire community.
Emma's Zero-Infrastructure Privacy: GDPR Without Extra Processes
Follow Emma, a privacy engineer at a European fintech, as she implements GDPR compliance without adding a single new process or service. Watch how mloda, running as a simple Python module in her existing infrastructure, transforms privacy compliance from architectural complexity into import statements.
Emma faces a dual challenge: implement GDPR compliance for millions of daily transactions, but the CTO declares 'no new services, no new infrastructure.' Traditional solutions require privacy gateways, audit services, and compliance APIs. Emma discovers that mloda runs as a Python module directly in their existing stack. That means no extra processes, no network calls, just Python imports that transform compliance from infrastructure burden to code elegance.
Emma implements comprehensive GDPR logging using mloda's Extender. It runs in-process with her existing Python application. No audit microservice to deploy, no message queues to configure, no network latency. The audit trail writes to her existing logging infrastructure. Every transformation is logged at Python speed, not API speed.
Emma's GDPR-compliant system is live, processing millions of transactions. The infrastructure team is amazed: zero new containers, zero new services, zero additional processes. Privacy controls, audit logging, and data minimization all run as Python modules in their existing application. The auditor's verdict: 'This is the simplest compliant architecture I've seen.' The CTO's verdict: 'No infrastructure complexity added!'
Get in touch
Questions about mloda, the roadmap, or working with us? Write to [email protected] or use the form.