Building muse — a private multi-model AI tool
Why I built muse: a BYOK AI tool that queries multiple LLMs, synthesizes one answer with a trust score, and runs on-device via MLX. Now with an iOS app.
All Posts
Writing about data engineering, AI, and things I'm building
Building a Modern Data Pipeline: Scala 3 + Kafka + Flink + Spark
How I built a streaming data pipeline using Scala 3 and Scala 2.13 together — with Kafka, Flink, Spark, and ZIO orchestration.
Building a Second Brain with LLMs
How I built a working knowledge graph of my own work in Obsidian using Andrej Karpathy's LLM Wiki approach.
slidesmd — Make Your PowerPoint Presentations Queryable by AI
Auto-index your .pptx files into a single agents.md that any AI can read. Drop a file in the folder, agents.md updates automatically. Ask Claude or a local Ollama model anything across all your decks.
ClipAI CLI — Local AI Text Assistant for the Terminal
A terminal-native AI text assistant powered by Ollama. No API key, no cloud — runs entirely on your machine with clipboard, pipe, and interactive TUI modes.
Optimizing Spark Timestamp Columns in Parquet Files
How to improve compression and query performance for high-cardinality timestamp columns in Apache Spark by switching from INT96 to INT64 encoding
Effective use of Spark driver
.. and stop crushing it!
Schema Evolution in Databricks Delta Lake
Adapting to changing data...
Deequ - An Open Source Data Quality Library
unit testing the data...
Spark Chained Transformations
In typical data engineering tasks, we often work with procedural style code for data transformations