Issue #334 - The ML Engineer 🤖
Alejandro Saucedo
AI & Tech Executive @ Zalando | Advisor @ UN, EU, ACM, etc | Join 70k+ ML Newsletter
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 for free at https://ethical.institute/mle.html ⭐
If you like the content please support the newsletter by sharing with your friends via ✉️ Email, 🐦 Twitter, 💼 Linkedin and 📕 Facebook!
This week in Machine Learning:
If you're looking for an interesting career opportunity, I'm hiring for a few roles including Data Science Manager (Forecasting), as well as Data Scientist (Forecasting) - check them out and please do share with your network!
LegoGPT is the right type of AI creativity we want to see; CMU researchers built a model designed to generate physically stable and buildable LEGO designs from text prompts through a fine-tuned LLaMa model: This is a really creative new model from Carnegie Mellon University which generates LEGO structures incrementally from text prompts - the implementation is quite interesting as it ensures each brick is valid and collision-free through brick-by-brick rejection sampling. Basically what this means is that in order to maintain physical stability it needs to integrate physics-aware rollbacks which enables reverting when running into an unstable design. It is also interesting to see the tokenization approach for representation of LEGO designs, and overall this seems like an interesting foundation for more relevant work to come which potentially can go beyond just LEGO.
What if we can use GenAI to improve our accents in foreign languages? And how do these models represent accents in their latent space? This is a fascinating research effort which captures language accents in Latent Spaces, which then allows for the ability to "correct" the accent towards native-sounding audio that can help learners to improve their accents. The samples provided make it quite intuitive on how these type of models can disrupt the language learning space, as they can be implemented for an individual specifically, highlighting the areas of highest potential improvement, and providing key phonetic areas with most divergence. The language learning space is certainly ripe for disruption as these type of innovations evolve, going beyond purely just audio.
Throughout the last few weeks we've continued to see new multi-modality GenAI models which support image operations, and this is a great comprehensive analysis of the top models on their compared performance: This qualitative benchmark provides a random set of "challenges" that are run across multiple of the most prominent GenAI models (FLUX, Gemini, ChatGPT, etc) and are compared based on their required iterations and final adherence. On the results, OpenAI 4o led with 9/11 successful prompts (often on the first or second try), Imagen 3 followed at 7/11, FLUX.1 and Gemini Flash 2.0 tied at 4/11, HiDream-I1 managed 3/11, and Midjourney v7 only 2/11. This is far from a formal / quantitative benchmark but it's interesting to see an intuitive comparison that confirms some personal observations when interacting with these.
Google’s research on measuring tech debt shows that effectively managing technical debt in production ML hinges on three pillars: 1) use engineer-driven surveys to categorize the ten most common debt types (e.g., migrations, testing gaps, stale docs) and focus only on those that actually hinder productivity; 2) complement—but don’t replace—these qualitative insights with targeted metrics (e.g., churn, TODOs) which alone have low recall; 3) adopt a four-level maturity model—reactive, proactive, strategic, structural—to institutionalize deliberate, visible trade-offs between speed and quality, ensuring that debt is incurred, tracked, and paid down as transparently as model accuracy or SLOs.
Diving into PyTorch’s internals can provide practitioners with meaningful intuition when tackling more complex and challenging ML projects at scale, and these are some important components: It is interesting to dive into some of the core components such as the underlying components that connect the Tensor with its underlying Storage that owns the raw memory, enabling zero-copy views via stride/offset manipulation, etc. A layer deeper we can dive into the C++ which code is organized into Python bindings, frontend engine and JIT, core kernels, and the core pytorch library components. Definitely relevant to check it out and dive deeper into the codebase, who knows perhaps this can even lead to a first pull request if you stumble across something relevant.
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.
Upcoming conferences where we're speaking:
Other upcoming MLOps conferences in 2025:
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.
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:
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.
Co-Founder, Head of Product
3dtech debt sounds tough, but tackling it is where real innovation happens. what gems have you found along the way?