Learn Data With Mark
I remember reading a rather technical article about Eddie Van Halen’s guitar rig in the French trade press back in the early ’80s that began: “Edward Van Halen is a fascinating little guy” (Edward Van Halen est un petit mec passionnant). I’d say the same thing today about Mark Needham and his YouTube channel: Learn Data With Mark. His videos are real gems of meticulously crafted content that deserve much wider exposure.
Mark Needham is undoubtedly a seasoned developer and an exceptional multimedia technician, as evidenced by the topics he covers and the remarkable level of detail in his presentations. His chosen video format is refreshingly concise, typically around five minutes long. Mark proves that with a carefully prepared script, you can deliver a wealth of useful information in a short span. This is a stark contrast to the endless ramblings of some YouTubers, who pad their content with repetitive or self-indulgent commentary. Mark’s approach exemplifies the principles of “LIM” (less is more) and to a certain extent “SIB” (small is beautiful) as well.
He is, without a doubt, a perfectionist. The lighting, the minimalist black background, and the precise presentation of code snippets all reflect his meticulous attention to detail. Code snippets are displayed and scrolled with flawless timing, enhancing both clarity and engagement. From what I can deduce, one of the tools he employs is asciinema, a command-line utility that records terminal output for playback during video recording. This ensures perfect synchronization between displayed code and commentary. Additionally, Mark seems to use some form of remote control (occasionally visible in his right hand) to manage playback of his pre-recorded scripts seamlessly.
Mark’s fondness for the IPython REPL is evident. This tool allows him to break down and execute code step by step, making his tutorials both didactic and engaging. His early videos focused on data analysis topics with tools like Pandas, DuckDB, Apache Parquet, and Apache Kafka. Over time, his channel has expanded to include advancements in AI, covering multimodal LLMs, ASR, Hugging Face, RAG, OpenAI, Qdrant, Ollama, Vector Search, and more.
Mark’s channel is relatively new, with the first video posted on August 18, 2022—just a few months before the release of ChatGPT. By July 2023, he began integrating AI and LLM topics into his videos, quickly offering tutorials on running open-source LLMs locally using frameworks like llama.cpp. These frameworks enable fast inference on consumer-grade hardware, with or without GPUs. Importantly, his tutorials address real-world problems and practical applications rather than serving as superficial demos.
For example, Mark often discusses methods for constraining LLMs to provide structured outputs from disparate datasets—a crucial capability in real-world development. Over time, I’ve noticed that he uses additional on-the-side tools that enhance developer productivity such as Poetry and, more recently, uv. All the code he demonstrates is freely available on his GitHub repo or his blog, and he makes a commendable effort to respond to viewer questions.
In short, Mark Needham is a “fascinating little guy” in the world of programming and AI education. His channel is a treasure trove for anyone interested in programming and the incredible possibilities offered by new AI-based tools. I highly recommend checking out his content—you’ll likely come away both informed and inspired.