Most of the jobs my orchestrator hands to Codex are small ones that do not need a frontier model, so I wired the Codex harness to a Qwen MoE running on my own Mac through LM Studio. The wiring was less obvious than expected, and this blogpost collects everything that broke on the way to a working setup, plus the DSPy-style loop it makes affordable.
For WaySightAI I set the constraint that raw time-series data never leaves the user's browser, which meant the CSV parsing, cleaning, stationarity tests, and forecasting all had to run client-side. The core math is written in Rust and compiled to WebAssembly, and this blogpost goes through the architecture, the benchmark numbers, and the tradeoffs.
A forecasting system where an LLM agent runs the whole lifecycle: exploratory analysis in a Code Interpreter sandbox, feature engineering, Bayesian hyperparameter tuning and training on SageMaker, and deployment behind a live endpoint. This blogpost goes through the three-layer architecture and the patterns that keep LLM-generated code contained.
My text-to-SQL agent handled single questions and got lost on follow-ups, so I fine-tuned Qwen 3.5 9B on multi-turn data and compared it with Claude Sonnet 4.6 on the same rows and scorer. The runs point at table and column selection, rather than SQL writing, as the real bottleneck for small models.
A chatbot that answers questions about the Strands SDK by searching its own source code: FAISS for retrieval, mem0 for conversation memory, Gemini for generation. Retrieval was done in a day while context assembly filled the rest of the week, and this blogpost walks through both, including the re-ranking corner I cut.
Data quality is a rather critical part of any production data pipeline. In order to provide accurate SLA metrics and to ensure that the data is correct, it is important to have a way to validate the data and report the metrics for further analysis.
Have you ever stumbled upon a Spark ETL and you were left wondering how a simple loading of a dataset can take hours, even though the filtered dataset you are specifying is relatively small?
In this blogpost we will continue our journey of testing our Data Pipelines. If you haven't checked out the first post, make sure you do.
Unit testing is often regarded as a main pillar of testing your software applications, and it usually involves testing a single/unit component and ensuring that it covers all the edge cases the software developer can think of.