🦉DVC

The Swiss AI Center (@hes_so) turned a full MLOps course into a space mission: DVC versions your data at liftoff, CML reviews model diffs in orbit, and drift alerts trigger reentry back into Label Studio for retraining. 👉🏽 hubs.la/Q04nYNL00

Muhammad Ishfaq trained a great emotion-detection model, then couldn't reproduce it. His fix: MLflow's registry for model versions, DVC for the data behind them, and treating experiment logging as non-negotiable from day one. 👉🏽 hubs.la/Q04nYX3X0

Sachin and Sameeksha Garg walk through a battle-tested MLOps stack for fraud detection in finance: DVC for file-level model versioning, blue-green @Kubernetes rollouts, and drift monitoring to catch when production data stops matching training data. 👉🏽 hubs.la/Q04nZ5cX0

When a "Data Scientist" says: "I know exactly what hyperparameters gave us the best model." DVC:

The Periodic Table of Agent Infrastructure lays out the building blocks of a modern agentic system More than 100 companies, organized the way the periodic table organizes elements. Explore the interactive version at hubs.la/Q04pRJ6R0

This AI agents course builds a production DVC registry for RAG knowledge bases and eval corpora, with point-in-time feature retrieval so an agent replay always sees the exact data state it saw the first time. 👉🏽 hubs.la/Q04nYhgh0

Forecast Studio treats time-series forecasting as engineering, not modeling: a five-stage pipeline (ingest, transform, train, evaluate, publish) with DVC versioning every artifact and a backtesting harness gating what ships. 👉🏽 hubs.la/Q04nYdfs0

@AWS built two DVC + SageMaker @MLflow lineage patterns — one traces patient records through a consent registry so opt-outs are provably excluded from retraining. 👉🏽 hubs.la/Q04nZwm80

A walkthrough of the OWASP AI Maturity Assessment across all 8 domains, using DVC for data provenance alongside fairlearn, SHAP, OPA, and garak to turn AI governance from checklist into actual tooling. 👉🏽 hubs.la/Q04nYM110

You can now use @huggingface Buckets as a @DVCorg remote. (DVC is a "git for data" tool, i.e. version datasets & models without stuffing large files into git.) Buckets speak S3 now → DVC points right at one: pointers in git, data in your bucket, no egress on pulls.

🚀 New: dvc-to-lakefs Zero-copy, one-command import from DVC into lakeFS. No re-uploading, it references the objects you already pushed. For when a project outgrows the laptop-and-cache rhythm. 👉🏽 hubs.la/Q04n21Mg0

We're co-hosting a webinar with @lakeFS to showcase how CytoReason are using @nextflowio, Seqera, and lakeFS to ship reproducible disease models as governed, versioned data products. Register now: hubs.la/Q04lpHJk0

lakeFS for Agentic AI Isolates and Reproduces Enterprise Data for Every Agent dbta.com/Editorial/News… #AgenticAI

@Moksh1326 @NamanBansal0611 @manicule Great work @Moksh1326 !! Keep learning and growing! 💪🏼

@NamanBansal0611 @manicule 2. Git and why it fails for ML systems (This blog was recognized by dvcorg itself) mokshmannu.substack.com/p/data-version…

We're hiring for a very specific role at @manicule. Call it Content Engineer? Essentially, I'm looking for someone who has a clear viewpoint on what good technical writing looks like. And, you're also comfortable building AI systems around it. You're a writer and an engineer. - Our job is to distill our human knowledge into a replicable system. - If there was one quality that mattered most, it would be attention to detail/obsession with quality. - Remote works, salary upto $200K/yr based on location. We rarely hire and stick with our team for long. Pls email me at naman@manicule[dot]dev Happy to pay $10K for successful referrals. Not using the usual buzzwords like high agency, taste, etc. that's a given. PS: I throw obviously AI applications in the bin.

"The manual model breaks": What happens when agents write to production data thenewstack.io/lakefs-agentic…

Mohd Azhar traces how ML went from handcrafted models to industrial pipelines with DVC, MLflow, and orchestration tools at the center. Real examples from Netflix, Uber, and Spotify, plus a full fraud detection pipeline walkthrough. 👉🏽 hubs.la/Q04fSvyy0

Have you all checked out @oboelabs? A solid production ML systems guide covering batch vs. stream ingestion, ETL/ELT patterns, data quality validation, and training-serving skew, with DVC and LakeFS recommended for data versioning and reproducibility. 🐇 hubs.la/Q04fSkwX0

Mehrdad Dehghan walks through a full end-to-end SaaS lead repurchase prediction pipeline, including EDA, LightGBM + Optuna tuning, DVC for pipeline orchestration, MLflow for tracking, and FastAPI deployment. Good stuff! 🚀 hubs.la/Q04fSyz_0

Lena Müller argues that winning AI teams won't have the biggest models. Instead, they'll have the most sophisticated evaluation pipelines. A deep dive into production-grade eval systems with DVC, CI/CD, and continuous monitoring. 👉🏽 hubs.la/Q04fQW0Q0

implemented dvc today but not in production tomorrow might just create an aws server and handle the dvc there in prod will continuee thisss atpp i should just start writing blogs for folks trying to get into ml beyond train_test_split and model.fit

Data Version Control in MLops so with git tracking our entire codebase we have a cap of size of files that we can track with help of git well if we talk about data in real world its volume is much higher than that cap so we use dvc which generates a sha id