I'm Emre, a software engineer with a background in Machine Learning and Computer Vision.

I'm having fun (and earning some money) with computers for more than a quarter century now. My first really useful project was in Delphi 1.0, around 1996. Before that I had some experience with Basic and C.

These days I'm working in Rust and Python. My last project was to find ships with Distributed Acoustic Sensor data. I built end-to-end training and deployment pipelines for large amount of data.

Before that I worked in LLM pyschology and persuasion to help lawyers in their contracts and legal documents. Earlier, I wrote techical documents for DVC, created end-to-end systems for face recognition, OCR for Ottoman Turkish and video OCR for Arabic.

Between my employments, I'm building an open source project Xvc, writing its docs and its Python bindings to manage large data and deep pipelines for Machine Learning projects. Install it with cargo install xvc for CLI and pip install xvc for Python. These days I'm also building an inboxbot as my agentic manager. Finally I'll have a perfect manager.

• Data is more important than the code: The goal of software is transforming data from one format to the other.

• Data is more important than hyperparameters: Having good quality data has more positive affect in ML model performance than model parameters.

• Functional programming leads to better maintanence: The more context you wrap around a piece of code to function, the more bugs you have. Classical OOP is not a good idea because of this.

• Incremental development over rearchitecting whole code bases: There is almost no reason to rewrite a piece of software from ground up.

• Code is a liability, not an asset: The less lines are used to solve a problem, the better.

• Slow is a code smell: If the code is prohibitevly slow to run and develop, there is something waiting to be understood about it.

Split instead of rewrite: There is a reason for a software project to go astray. Trying to evolve by splitting functionality leads to understanding those reasons instead of blind rewrites.

• Thinking types is time well-spent: In the long run, for anything serious you'll have to make decisions with types. This will make you understand better.

• Keyboard over mouse: If you cannot automate it with a keyboard, it's not a good design.

• Growing forests is a better analogy for software then building: Especially with agentic development, growing strong and healthy forests on data soil is a better analogy for software than building concrete skyscrapers.

• Writing is better than talking to think: Maybe it's my slow mind, but I find writing on the same topic over and over a better approach when finding solutions than meetings.

• Mistakes are better than perfection: Accepting you'll never have the perfect solution is a good start to find a feasible one.

• Small teams are more productive: When working with larger problems, adding more people increases communication friction. If you cannot solve a problem with 5 people, it probably requires a rethinking and replanning on the whole.

I live in Istanbul and Mudanya. You can get in touch with me via email (contact of emresahin.net) or Telegram emre_sahin.