kongastral
Software engineer working in machine learning systems, distributed data infrastructure, and back-end engineering. AI Code Invest is the author’s technical notebook, made public.
About AI Code Invest
AI Code Invest is a technical publication focused on two adjacent areas that the author works in daily: applied artificial intelligence and machine learning, and software and data engineering. Articles are written for engineers and researchers who need to implement, evaluate, or troubleshoot a system, and who prefer a careful, source-grounded treatment to a marketing summary.
The name reflects the site’s purpose rather than a financial subject: it is about investing time and effort into AI and code — that is, the deliberate, long-term practice of building engineering and machine-learning skill. The site does not cover stock-market investing or any other financial topic, and it does not provide financial, medical, legal, or other professional advice. It publishes only in the two technical areas named above.
About the author
Hi — I’m kongastral. I work as a software engineer focused on machine learning infrastructure, time-series analytics, and back-end systems. My day-to-day involves building production data pipelines, integrating ML models into operational systems, and reviewing research that may translate into engineering practice.
I started this site because I kept writing the same internal documents at work — design notes on Kafka topology, post-mortems on data pipeline failures, summaries of new papers I had read, walk-through guides for new team members. Many of those notes were useful beyond my immediate team. AI Code Invest is the public version of that notebook.
What I write about
- Applied AI and ML: anomaly detection (one-class methods such as Deep SVDD, OCSVM), domain adaptation (DANN and related adversarial methods), self-supervised and semi-supervised learning (SimCLR, MAE, FixMatch and others), Gaussian Processes, time-series modelling, large-language-model engineering, and reinforcement-learning-adjacent topics where applicable.
- Software and data engineering: Apache Kafka, Apache Airflow, Debezium change data capture, dbt, Docker, Kubernetes, Python, FastAPI, and the broader infrastructure required to run analytic and ML workloads in production.
How I write
Articles are researched, drafted, fact-checked, and edited by me. For each post:
- I read the primary references — the original paper, official documentation, or source code — before drafting.
- I run every code sample locally and check the output. Snippets that do not produce the stated result are corrected or removed.
- I verify every external link by visiting it and confirming it supports the claim.
- I use the same writing tools any technical author uses today: editors, spell checkers, grammar checkers, and reference managers.
- I do not publish content I have not personally verified.
The full editorial process is described in the Editorial Standards page.
Why this site exists
Three motivations:
- To consolidate what I learn. Writing a topic up forces a deeper understanding than reading. Many of these articles started as notes I wrote for myself.
- To make practical material accessible. Much technical content online is either too shallow (tutorial-style) or too dense (paper-style). I try to occupy a middle ground that gives an engineer enough depth to make real decisions without obscuring the underlying ideas.
- To document working systems. Many of the patterns described here have been used in production by myself or by colleagues. Where that is the case, the article says so.
What this site is not
- It is not a paid newsletter. There is no subscription tier.
- It is not a consulting or advisory service.
- It is not affiliated with any vendor, library author, or company whose products are discussed in articles. Mentions and recommendations are made on technical merit.
The site displays Google AdSense advertising, which covers hosting and operating costs. Advertising decisions are made by the ad network and are not influenced by editorial considerations.
Contact
For questions, corrections, topic suggestions, or any other reason to reach the author, please write to kongastral@gmail.com. Replies usually arrive within a few working days. Constructive corrections from readers are particularly welcome — the Editorial Standards page describes how they are handled.
Privacy
For information on how visitor data is handled, please see the Privacy Policy.
Last updated: June 2026.