THE INSTITUTE FOR ETHICAL AIALIGNMENT + SAFETY
JOIN / CONTACT

THE ML ENGINEER — WEEKLY NEWSLETTER

The MachineLearning EngineerIssue #369

Join 70k+ AI professionals receiving weekly curatedarticles, tutorials and blog posts onproduction machine learning.

#368
Issue #36911/01/26llmsai-agentsdata-engineering
#370
Issue #369 🤖 - Databases 2025 Year in Review, Lessons from 21 Years at Google, Raschka's Big LLM Architectures, Measuring AI Ability for Long Tasks + more 🚀

Thank you for being part of over 70,000+ ML professionals and enthusiasts who receive weekly articles & tutorials on Machine Learning & MLOps 🤖 You can join the newsletter https://bit.ly/state-of-ml-2025 ⭐

If you like the content please support the newsletter by sharing with your friends via ✉️ Email, 🐦 Twitter, 💼 Linkedin and 📕 Facebook!

This week in ML Engineering:

Databases 2025 Year in Review

The Databases 2025 Year-in-Review from Carnegie Mellon University is out! As always there are some great insights on database trends across last year: As expected Postgres dominance continues to accelerating, particularly with “postgres-adjacent” products, like Databricks acquiring Neon ($1B), Snowflake acquiring Crunchy Data ($250M), and Microsoft launching a new Postgres DBaaS, together with many other new middleware projects (Multigres, Neki, PgDog). For production ML teams, the biggest operational shift is that databases are becoming “agent-facing” by design, with huge adoption of MCP as a standard way for LLMs to invoke database tools. There are some areas to keep an eye such as the API-compatibility as it’s turning into a legal minefield (MongoDB suing FerretDB), as well as open data formats having another re-surgence with multiple new columnar formats challenging Parquet and pushing new ideas. Databases are core to the day-to-day software operations, so we will likely see some exciting new develolpments in the year to come, especially with the craze on agentic development.

Lessons from 21 Years at Google

“21 Lessons From 14 Years at Google” - the top 5 that resonated the most for me: 1) You can win every technical argument and lose the project. 2) Bias towards action. Ship. You can edit a bad page, but you can’t edit a blank one. 3) Your code doesn’t advocate for you. People do. 4) At scale, even your bugs have users. 5) Most “slow” teams are actually misaligned teams. This really is a great list as it reminds all of us how some of the most challenging and useful lessons are often not related to the code but to the soft skills and people interactions.

The State of MLOps 2025 Survey 🔥

The level of ML maturity of organisations in 2025 has slowly increased, and we have interesting insights on organisational departments/functions from our Prod ML Survey:

  • In 2025 about 29% of organisations have set up a central machine learning platform team.
  • However only 27% organisations have established a data platform / data engineering organisation.
  • There is a significant increase with 17% organisations now establishing an AI Risk & Governance Function.
  • However only 9% of organisations have established an AI inventory to keep track of all use-cases and models in the organisations.

If you want to dive deeper you can access the full results here: https://ethical.institute/state-of-ml-2025 🔥

Raschka’s Big LLM Architectures

Sebastian Raschka has put together an overview on the evolution of LLM architectures, together with the most comprehensive set of intuitive visual diagrams. It is interesting to see that Sebastian Raschka argues that these models are still structurally similar to when they were initially released. Positional embeddings have evolved from absolute to rotational (RoPE), Multi-Head Attention has largely given way to Grouped-Query Attention, and the more efficient SwiGLU has replaced activation functions like GELU, however it is yet to be seen if there’s been step-change breakthroughs since inception. Definitely worth diving into what seems to be one of the most comprehensive (intutive) overviews of LLMs out there.

.

Measuring AI Ability for Long Tasks

METR has released an in-depth study that measures the ability for LLM agents to solve long tasks, and there’s quite a lot of interest insights: In this study they first propose a practical way to quantify agentic LLM capability using a task completion horizon, which involves the human time-to-complete where an AI agent succeeds with 50% probability. Across their datasets it shows that task duration strongly predicts reliability, with near-perfect success on very short tasks but steep drop-offs on hour-scale tasks. Although this is intuitive it is great to see some data, which also shows how frontier models has been following a robust exponential trend since 2019, doubling roughly every ~7 months. There are some interesting developments that attribute to this, including better tool use, reasoning, and recovery from mistakes rather than mere knowledge. This is a space that will only expand in 2026, with a lot of exciting developments ahead.

Upcoming MLOps Events

The MLOps ecosystem continues to grow at break-neck speeds, making it ever harder for us as practitioners to stay up to date with relevant developments. A fantsatic way to keep on-top of relevant resources is through the great community and events that the MLOps and Production ML ecosystem offers. This is the reason why we have started curating a list of upcoming events in the space, which are outlined below.

Conferences for 2026 coming soon! For the meantime, in case you missed our talks:

Open Source MLOps Tools

Check out the fast-growing ecosystem of production ML tools & frameworks at the github repository which has reached over 10,000 ⭐ github stars. We are currently looking for more libraries to add - if you know of any that are not listed, please let us know or feel free to add a PR. Four featured libraries in the GPU acceleration space are outlined below.

  • Kompute - Blazing fast, lightweight and mobile phone-enabled GPU compute framework optimized for advanced data processing usecases.
  • CuPy - An implementation of NumPy-compatible multi-dimensional array on CUDA. CuPy consists of the core multi-dimensional array class, cupy.ndarray, and many functions on it.
  • Jax - Composable transformations of Python+NumPy programs: differentiate, vectorize, JIT to GPU/TPU, and more
  • CuDF - Built based on the Apache Arrow columnar memory format, cuDF is a GPU DataFrame library for loading, joining, aggregating, filtering, and otherwise manipulating data.

If you know of any open source and open community events that are not listed do give us a heads up so we can add them!

OSS: Policy & Guidelines

As AI systems become more prevalent in society, we face bigger and tougher societal challenges. We have seen a large number of resources that aim to takle these challenges in the form of AI Guidelines, Principles, Ethics Frameworks, etc, however there are so many resources it is hard to navigate. Because of this we started an Open Source initiative that aims to map the ecosystem to make it simpler to navigate. You can find multiple principles in the repo - some examples include the following:

  • MLSecOps Top 10 Vulnerabilities - This is an initiative that aims to further the field of machine learning security by identifying the top 10 most common vulnerabiliites in the machine learning lifecycle as well as best practices.
  • AI & Machine Learning 8 principles for Responsible ML - The Institute for Ethical AI & Machine Learning has put together 8 principles for responsible machine learning that are to be adopted by individuals and delivery teams designing, building and operating machine learning systems.
  • An Evaluation of Guidelines - The Ethics of Ethics; A research paper that analyses multiple Ethics principles.
  • ACM’s Code of Ethics and Professional Conduct - This is the code of ethics that has been put together in 1992 by the Association for Computer Machinery and updated in 2018.

If you know of any guidelines that are not in the “Awesome AI Guidelines” list, please do give us a heads up or feel free to add a pull request!

About us

The Institute for Ethical AI & Machine Learning is a European research centre that carries out world-class research into responsible machine learning.

Check out our website

✉️ Email, 🐦 Twitter, 💼 Linkedin