The ML Engineer
LightNews LightNews
ml.blaze.email
The ML Engineer
@ml.blaze.email
9 followers 1 following 420 posts
Posts about ML engineering, MLOps and ML research (and a sprinkling of AI). Weekly newsletter digest at ml.blaze.email. Maintained by @alastairrushworth.com.
Posts Replies Media Videos
Pinned
ml.blaze.email
The ML Engineer @ml.blaze.email · Jan 13
The latest edition of The ML Engineer has dropped! Check out all the latest news and posts from the ML community in a single digest.

https://ml.blaze.email/archive/the-ml-engineer-13-01-2026/

#machinelearning #datascience
The ML Engineer 13-01-2026
🔧 Company Engineering Blogs Why We Use Separate Tech Stacks for Personalization and Experimentation (engineering​.atspotify​.com). Spotify separates ML...
ml.blaze.email
ml.blaze.email
microgpt

http://karpathy.github.io/2026/02/12/microgpt/

#machinelearning #datascience
microgpt
Summary: MicroGPT: a 200-line pure-Python autograd GPT-2-like trainer, tokenizer, RMSNorm, and Adam optimizer in one file
karpathy.github.io
February 14, 2026 at 12:13 PM
ml.blaze.email
Data is the new Lego

https://blog.engora.com/2026/02/data-is-new-lego.html

#machinelearning #datascience
Data is the new Lego
Summary: Data, Lego, and fashion AI: pooling diverse data, modeling, and recommendations with True Fit's Genome, using AI/ML concepts
blog.engora.com
February 14, 2026 at 6:13 AM
ml.blaze.email
Are AI SQL yet?

https://www.viblo.se/posts/are-ai-sql-yet/

#machinelearning #datascience
Are AI SQL yet?
Summary: Testing AI for SQL tasks on Azure SQL, highlighting safe downtime strategies and DATA_CONSISTENCY_CHECK settings for system-versioned tables.
www.viblo.se
February 14, 2026 at 12:13 AM
ml.blaze.email
Boxworld and AI Pipelining

https://ulyssepence.com/blog/post/boxworld-and-ai-pipelining

#machinelearning #datascience
Boxworld and AI Pipelining
Summary: Boxworld uses Python (gymnasium, stable_baselines3) to train an agent, then visualizes outcomes in React Three Fiber for web.
ulyssepence.com
February 13, 2026 at 6:12 PM
ml.blaze.email
Advanced RAG — Understanding Reciprocal Rank Fusion in Hybrid Search

https://glaforge.dev/posts/2026/02/10/advanced-rag-understanding-reciprocal-rank-fusion-in-hybrid-search/

#machinelearning #datascience
Advanced RAG — Understanding Reciprocal Rank Fusion in Hybrid Search
Summary: Hybrid search with RRF, BM25, vector retrievals, LangChain4j, and Java-based RAG architectures explained using k=60 and a cross-encoder two-stage approach
glaforge.dev
February 13, 2026 at 12:12 PM
ml.blaze.email
Hierarchical Navigable Small Worlds

https://amandeepsp.github.io/blog/hnsw/

#machinelearning #datascience
Hierarchical Navigable Small Worlds
Summary: Explains HNSW: Delaunay, NSW, Kleinberg rings, hierarchical graphs, and greedy search in Python-inspired terms.
amandeepsp.github.io
February 13, 2026 at 6:12 AM
ml.blaze.email
back to shrinkage!

https://xianblog.wordpress.com/2026/02/12/back-to-shrinkage/

#machinelearning #datascience
back to shrinkage!
Summary: Shrinkage priors for implicit generative models with high-dimensional parameters using block SGMCMC and generalized Bayes, applied to weather, RL, and linear regressions
xianblog.wordpress.com
February 13, 2026 at 12:12 AM
ml.blaze.email
The perceptron

https://blog.engora.com/2026/02/the-perceptron.html

#machinelearning #datascience
The perceptron
Summary: Overview of the perceptron, its theory, training rules, historical hype, and impact on AI development.
blog.engora.com
February 12, 2026 at 6:13 PM
ml.blaze.email
Outsourced Intelligence: The Hidden Attack Surface in Enterprise AI

https://kiledjian.com/2026/02/10/outsourced-intelligence-the-hidden-attack.html

#machinelearning #datascience
Outsourced Intelligence: The Hidden Attack Surface in Enterprise AI
Summary: DMaaS risk: backdoored AI decision models across enterprise systems, governance gaps, and need for provenance, drift monitoring, and independent paths.
kiledjian.com
February 12, 2026 at 12:13 PM
ml.blaze.email
Error and Bias Prediction Metrics for Regression

https://jmsallan.netlify.app/blog/error-and-bias-prediction-metrics-for-regression/

#machinelearning #datascience
Error and Bias Prediction Metrics for Regression
Summary: Regression error and bias metrics using RMSE, MAE, MAPE, SMAPE, MSD, MPE with tidymodels in R on the Ames dataset
jmsallan.netlify.app
February 12, 2026 at 6:13 AM
ml.blaze.email
PFT: Phonon Fine-tuning for Machine Learned Interatomic Potentials

https://teddykoker.com/2026/02/pft/

#machinelearning #datascience
PFT: Phonon Fine-tuning for Machine Learned Interatomic Potentials
Summary: Phonon fine-tuning for ML interatomic potentials using Hessian(V) training with Nequix in JAX and PyTorch
teddykoker.com
February 12, 2026 at 12:13 AM
ml.blaze.email
Adventures in Neural Rendering

https://interplayoflight.wordpress.com/2026/02/10/adventures-in-neural-rendering/

#machinelearning #datascience
Adventures in Neural Rendering
Summary: Kostas Anagnostou explores neural networks, MLPs, and SH representations for rendering tasks in a graphics programming context.
interplayoflight.wordpress.com
February 11, 2026 at 6:12 PM
ml.blaze.email
Building a semantic search engine in ±250 lines of Python

https://bart.degoe.de/building-a-semantic-search-engine-in-250-lines-of-python/

#machinelearning #datascience
Building a semantic search engine in ±250 lines of Python
Summary: Semantic search with sentence-transformers embeddings and cosine similarity in Python
bart.degoe.de
February 11, 2026 at 12:13 PM
ml.blaze.email
Architectural and Mathematical Foundations of Machine Learning: A Rigorous Synthesis of Theory, Geometry, and Implementation

https://chizkidd.github.io/2026/02/09/mathematical-machine-learning-foundations/

#machinelearning #datascience
Architectural and Mathematical Foundations of Machine Learning: A Rigorous Synthesis of Theory, Geometry, and Implementation
Summary: Rigorous synthesis of math foundations in ML: linear/affine geometry, SVD, information theory, optimization, probabilistic estimation, kernels, attention, diffusion, and robust numerics.
chizkidd.github.io
February 11, 2026 at 6:12 AM
ml.blaze.email
The Modern Python Data Stack in 2026

https://htdocs.dev/posts/the-modern-python-data-stack-in-2026/

#machinelearning #datascience
The Modern Python Data Stack in 2026
Summary: A guide to the modern Python data stack in 2026 featuring uv, Ruff, ty, Marimo, Polars, DuckDB, Positron, Quarto, and Evidence.
htdocs.dev
February 11, 2026 at 12:13 AM
ml.blaze.email
Getting to the bottom of TMLE: influence functions and perturbations

https://www.rdatagen.net/post/2026-02-05-getting-to-the-bottom-of-tmle-1/

#machinelearning #datascience
Getting to the bottom of TMLE: influence functions and perturbations
Summary: Explains TMLE concepts like influence functions, perturbations, and orthogonality using statistical theory and examples in causal inference.
www.rdatagen.net
February 10, 2026 at 6:13 PM
ml.blaze.email
Zero ETL i hackathon z Bielikiem

https://blog.prokulski.science/2026/02/09/zero-etl-i-hackathon-z-bielikiem/?utm_source=rss&utm_medium=rss&utm_campaign=zero-etl-i-hackathon-z-bielikiem

#machinelearning #datascience
Zero ETL i hackathon z Bielikiem
Summary: Zero ETL jako standard w inżynierii danych, praktyczne zastosowania AI w data science i ewolucja DuckDB w lakehouse z materiałami i wyzwaniami DWthon
blog.prokulski.science
February 10, 2026 at 12:13 PM
ml.blaze.email
Modernizing my 150-line Python search engine: Yahoo! dumps -> Hugging Face 🤗

https://bart.degoe.de/modernizing-python-search-engine/

#machinelearning #datascience
Modernizing my 150-line Python search engine: Yahoo! dumps -> Hugging Face 🤗
Summary: Modernizing a 150-line Python search engine using Hugging Face datasets, uv, ruff, pytest, and GitHub Actions
bart.degoe.de
February 10, 2026 at 6:13 AM
ml.blaze.email
Certified AI Developer - A Journey of Growth, Grief, and New Beginning

https://wirtel.be/post/2026/02/05/2026-certified-ai-developer/

#machinelearning #datascience
Certified AI Developer - A Journey of Growth, Grief, and New Beginning
Summary: Python developer with 20+ years learns ML/DL fundamentals using Alyra certification, covering data prep, regression, classification, neural networks, and deployment
wirtel.be
February 10, 2026 at 12:13 AM
ml.blaze.email
Buying Space, Not Keywords: A UX for Advertising in Embedding Space

https://www.june.kim/buying-space-not-keywords

#machinelearning #datascience
Buying Space, Not Keywords: A UX for Advertising in Embedding Space
Summary: Semantics-driven ad targeting in embedding space using semantic hill-climbing, dimensionality challenges, and geometry-based brand safety.
www.june.kim
February 9, 2026 at 6:13 PM
ml.blaze.email
AI@home: Classifying images with Ollama – part one.

https://vegard.blog.engen.priv.no/?p=607

#machinelearning #datascience
AI@home: Classifying images with Ollama – part one.
Summary: Explores image/video classification using Ollama with LLaVA models, FastAPI, Kubernetes, and Python tooling—including frame extraction and JSON prompts.
vegard.blog.engen.priv.no
February 9, 2026 at 12:13 PM
ml.blaze.email
Paths of MySQL, vector search edition

http://notes.eatonphil.com/2026-02-08-paths-of-mysql-vector-search-edition.html

#machinelearning #datascience
Paths of MySQL, vector search edition
Summary: Explores MySQL vector search workflows, Python tooling, and performance tips with practical examples and guidance.
notes.eatonphil.com
February 9, 2026 at 6:13 AM
ml.blaze.email
Training a 360M Parameter Model with Performance Discipline

https://www.saisasank.com/training-a-360m-parameter-model-with-performance-discipline/

#machinelearning #datascience
Training a 360M Parameter Model with Performance Discipline
Summary: Pretrain a 360M LLM on a single A100 in 30 hours, maximizing MFU with BF16 AMP, gradient accumulation, and data pipeline tuning.
www.saisasank.com
February 9, 2026 at 12:13 AM
ml.blaze.email
FOSDEM 2026. A weekend of tech talks and beer

https://valatka.dev/2026/02/07/fosdem-26.html

#machinelearning #datascience
FOSDEM 2026. A weekend of tech talks and beer
Summary: FOSDEM 2026 covers Rust, Arrow, lakehouses, type checkers, containers, and talks from RonDB to DuckDB and ClickHouse
valatka.dev
February 8, 2026 at 6:12 PM
ml.blaze.email
Build Better Strategies, Part 6: Evaluation

https://financial-hacker.com/build-better-strategies-part-6-evaluation/

#machinelearning #datascience
Build Better Strategies, Part 6: Evaluation
Summary: Automated evaluation framework in Zorro for strategy testing using shells, WFO, Monte Carlo, and robust parameter analysis.
financial-hacker.com
February 8, 2026 at 12:12 PM