The ML Engineer
LightNews LightNews
ml.blaze.email
The ML Engineer
@ml.blaze.email
8 followers 1 following 220 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 · 2d
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-23-12-2025/

#machinelearning #datascience
The ML Engineer 23-12-2025
ml.blaze.email
ml.blaze.email
Predicting survival using a super learner and right-censored data

https://aliceinstatisticsland.wordpress.com/2025/12/21/predicting-survival-using-a-super-learner-and-right-censored-data/

#machinelearning #datascience
Predicting survival using a super learner and right-censored data
Summary: Survival analysis with a super learner using right-censored data in R (survivalSL, flexsurv, glmnet) and methods like randomSRC and survival neural networks
aliceinstatisticsland.wordpress.com
December 26, 2025 at 12:13 AM
ml.blaze.email
Property-based testing, adversarial developers, and LLMs

https://protocols-made-fun.com/pbt/2025/12/22/pbt-adversarial-llms.html

#machinelearning #datascience
Property-based testing, adversarial developers, and LLMs
Summary: Property-based testing, adversarial developers, LLMs, Hypothesis, Apalache, and Python-based examples in addition, commutativity, and associativity checks
protocols-made-fun.com
December 25, 2025 at 6:13 PM
ml.blaze.email
Initial Quick Thoughts on Singular Learning Theory

http://www.beren.io/2025-12-24-Initial-Quick-Thoughts-on-Singular-Learning-Theory/

#machinelearning #datascience
Initial Quick Thoughts on Singular Learning Theory
Summary: Explores singular learning theory (SLT) concepts, posterior geometry, WBIC, SGD noise, and practical implications for generalization and optimization.
www.beren.io
December 25, 2025 at 12:12 PM
ml.blaze.email
Choosing the Right LLM for Cognee: Local Ollama Setup

https://www.glukhov.org/post/2025/12/choosing-right-llm-for-cognee-on-ollama/

#machinelearning #datascience
Choosing the Right LLM for Cognee: Local Ollama Setup
Summary: Local Ollama setup for Cognee using Devstral, qwen3:14b, and other LLMs with embedding strategies and low-hallucination trade-offs.
www.glukhov.org
December 25, 2025 at 6:13 AM
ml.blaze.email
Modern-AI Benchmarking – which models? which benchmarks? what use cases?

https://blog.davidi.com/2025/12/21/modern-ai-benchmarking-which-models-which-benchmarks-what-use-cases/

#machinelearning #datascience
Modern-AI Benchmarking – which models? which benchmarks? what use cases?
Summary: Explores AI benchmarks, use cases, and leading boards (MLCommons, Epoch.ai, LiveBench, BetterBench, LLM Leaderboard) for modern LLMs and AI systems.
blog.davidi.com
December 25, 2025 at 12:13 AM
ml.blaze.email
Machine Learning in Clojure with libpython-clj: Bridging Functional Elegance and Python’s ML Power [Series 1]

https://flexiana.com/hidden/2025/12/machine-learning-in-clojure-with-libpython-clj-bridging-functional-elegance-and-pythons-ml-power-series-1

#machinelearning #datascience
Machine Learning in Clojure with libpython-clj: Bridging Functional Elegance and Python’s ML Power [Series 1]
Summary: Bridge Python’s ML power with Clojure via libpython-clj to train in Python and infer in the JVM.
flexiana.com
December 24, 2025 at 6:13 PM
ml.blaze.email
Machine Learning in Clojure with libpython‑clj: Unlocking Causal Insights Using Microsoft’s EconML [Series 3]

https://flexiana.com/news/2025/12/machine-learning-in-clojure-with-libpython-clj-unlocking-causal-insights-using-microsofts-econml-series-3

#machinelearning #datascience
Machine Learning in Clojure with libpython‑clj: Unlocking Causal Insights Using Microsoft’s EconML [Series 3]
Summary: Using libpython-clj to leverage EconML for causal inference in Clojure with practical insights on targeting and heterogenous effects.
flexiana.com
December 24, 2025 at 12:13 PM
ml.blaze.email
Eine [Studie von Nature Communications](https://www.nature.com/articles/s41467-025-67728-y) besch...

https://mikka.is/posts/prairie-cloud-715/

#machinelearning #datascience
Eine [Studie von Nature Communications](https://www.nature.com/articles/s41467-025-67728-y) besch...
Summary: Portable device for frailty assessment using high-rate gait biosignals KI inference; continuous 10-day operation; automated ward vital-sign alerts for early deterioration; passive long-COVID sensing with ML.
mikka.is
December 24, 2025 at 6:13 AM
ml.blaze.email
Speculative Decoding in LLMs

https://uditsaxena.bearblog.dev/spec-decode/

#machinelearning #datascience
Speculative Decoding in LLMs
Summary: Speculative decoding techniques for LLMs, latency gains, vLLM and SGLang systems, and performance tradeoffs.
uditsaxena.bearblog.dev
December 24, 2025 at 12:13 AM
ml.blaze.email
Do a sanity check on your experiments

https://ehudreiter.com/2025/12/22/do-a-sanity-check-on-your-experiments/

#machinelearning #datascience
Do a sanity check on your experiments
Summary: Sanity checks on data, model outputs, and evaluation to detect bugs in NLP/AI experiments
ehudreiter.com
December 23, 2025 at 6:13 PM
ml.blaze.email
The Illustrated Transformer

https://jalammar.github.io/illustrated-transformer/

#machinelearning #datascience
The Illustrated Transformer
Summary: Overview of The Transformer architecture with self-attention, multi-head attention, positional encoding, and training dynamics using TensorFlow/T2T and PyTorch guides.
jalammar.github.io
December 23, 2025 at 12:13 PM
ml.blaze.email
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-23-12-2025/

#machinelearning #datascience
The ML Engineer 23-12-2025
ml.blaze.email
December 23, 2025 at 9:11 AM
ml.blaze.email
Machine Learning in Clojure with libpython‑clj: Unlocking Causal Insights Using Microsoft’s EconML [Series 3]

https://flexiana.com/hidden/2025/12/machine-learning-in-clojure-with-libpython-clj-unlocking-causal-insights-using-microsofts-econml-series-3

#machinelearning #datascience
Machine Learning in Clojure with libpython‑clj: Unlocking Causal Insights Using Microsoft’s EconML [Series 3]
Summary: Clojure + libpython-clj enable EconML causal inference, showing heterogeneous treatment effects beyond A/B testing.
flexiana.com
December 23, 2025 at 6:13 AM
ml.blaze.email
2025 in review

https://veekaybee.github.io/2025/12/22/2025-in-review/

#machinelearning #datascience
2025 in review
Summary: Production-ready ML systems: tokenization, error handling, Dockerization, logging, and the human craft behind building in public
veekaybee.github.io
December 23, 2025 at 12:13 AM
ml.blaze.email
Trying to fit a logistic curve

https://www.johndcook.com/blog/2025/12/20/fit-logistic-curve/

#machinelearning #datascience
Trying to fit a logistic curve
Summary: Fitting a logistic curve from left-tail data with Python's SciPy may fail or be imprecise
www.johndcook.com
December 22, 2025 at 6:13 PM
ml.blaze.email
[译][论文] P5 paper | 用语言模型做推荐：一种统一的预训练、个性化提示和预测范式（2022）

https://arthurchiao.github.io/blog/p5-paper-zh/

#machinelearning #datascience
[译][论文] P5 paper | 用语言模型做推荐：一种统一的预训练、个性化提示和预测范式（2022）
Summary: Unified pretraining with personalized prompts and text-to-text generation for recommender systems using P5 (2022) on multi-task datasets.
arthurchiao.github.io
December 22, 2025 at 12:13 PM
ml.blaze.email
Embedding-Based Tool Selection for AI Agents

https://zarar.dev/embedding-based-tool-selection-for-ai-agents/

#machinelearning #datascience
Embedding-Based Tool Selection for AI Agents
Summary: Embedding-based tool selection using pgvector in Postgres, OpenAI embeddings, and category expansions to scale AI agents’ tools with Elixir code.
zarar.dev
December 22, 2025 at 6:12 AM
ml.blaze.email
Sam Rose explains how LLMs work with a visual essay

https://simonwillison.net/2025/Dec/19/sam-rose-llms/#atom-everything

#machinelearning #datascience
Sam Rose explains how LLMs work with a visual essay
Summary: Visual explainer shows how LLMs work, covering tokenization, embeddings, prompt caching, and transformer basics with accessible diagrams and examples.
simonwillison.net
December 22, 2025 at 12:13 AM
ml.blaze.email
Construcción de intervalos de confianza para gráficos de calibración vía "bootstrap" y algunos asuntos más

https://datanalytics.com/2025/12/18/cortos/

#machinelearning #datascience
Construcción de intervalos de confianza para gráficos de calibración vía "bootstrap" y algunos asuntos más
Summary: Calibración de gráficos con bootstrap, intervalos de confianza y temas relacionados en estadística y ML usando R y Python.
datanalytics.com
December 21, 2025 at 6:13 PM
ml.blaze.email
La Rivoluzione dell'IA: Come i Modelli di Diffusione e Nuove Architetture Sostituiranno gli LLM Attuali

https://grigio.org/la-rivoluzione-dellia-come-i-modelli-di-diffusione-e-nuove-architetture-sostituiranno-gli-llm-attuali/

#machinelearning #datascience
La Rivoluzione dell'IA: Come i Modelli di Diffusione e Nuove Architetture Sostituiranno gli LLM Attuali
Summary: Diffusion models, sub-quadratic architectures, private thinking, continuous learning, and a continuous-thinking machine reshape AI beyond current LLMs.
grigio.org
December 21, 2025 at 12:13 PM
ml.blaze.email
Accelerating Tree-Based Models in SQL with Orbital 0.3.0

https://posit.co/blog/orbital-0-3-0/

#machinelearning #datascience
Accelerating Tree-Based Models in SQL with Orbital 0.3.0
Summary: Orbital 0.3.0 accelerates SQL-based tree models using Posit tools for R, Python, and SQL environments with centralized management and sharing capabilities.
posit.co
December 21, 2025 at 6:13 AM
ml.blaze.email
Advent of 2025, Day 16 – SQL Server 2025 – Optional parameter plan optimization (OPPO)

https://tomaztsql.wordpress.com/2025/12/16/advent-of-2025-day-16-sql-server-2025-optional-parameter-plan-optimization-oppo/

#machinelearning #datascience
Advent of 2025, Day 16 – SQL Server 2025 – Optional parameter plan optimization (OPPO)
Summary: SQL Server 2025 OPPO optional parameter plan optimization explores parameter-sniffing mitigation and adaptive plans with IXP/PSPO concepts
tomaztsql.wordpress.com
December 21, 2025 at 12:13 AM
ml.blaze.email
FractalNet Causal de Baja Entropía: La Arquitectura de la Inteligencia Soberana

https://juandomingofarnos.wordpress.com/2025/12/19/fractalnet-causal-de-baja-entropia-la-arquitectura-de-la-inteligencia-soberana/

#machinelearning #datascience
FractalNet Causal de Baja Entropía: La Arquitectura de la Inteligencia Soberana
Summary: FractalNet causal de baja entropía para AGI y educación disruptiva; ARC CT, JEPA-Causal, DOIs, LOGOS, Python-like pseudocódigo, Juan Domingo Farnós.
juandomingofarnos.wordpress.com
December 20, 2025 at 6:12 PM
ml.blaze.email
Scott Haines on the Future of Data Engineering

https://www.confessionsofadataguy.com/scott-haines-on-the-future-of-data-engineering/

#machinelearning #datascience
Scott Haines on the Future of Data Engineering
Summary: Scott Haines discusses the future of data engineering, cloud-native tooling, and practical approaches with Python, SQL, SQLAlchemy, and modern data stacks.
www.confessionsofadataguy.com
December 20, 2025 at 12:13 PM
ml.blaze.email
Measuring AI’s Impact on Shipping Speed and Code Quality

https://elezea.com/2025/12/measuring-ais-impact-on-shipping-speed-and-code-quality/

#machinelearning #datascience
Measuring AI’s Impact on Shipping Speed and Code Quality
Summary: Explores measuring AI impact on shipping speed and code quality using Python, benchmarks, and metrics in software development workflows.
elezea.com
December 20, 2025 at 6:13 AM