I’m Ven Petrov.

What I’m building and exploring now

I spend a lot of time working with LLMs and tools like Codex, Claude Code, OpenCode and Pi Agent. I build prototypes that connect models to data, tools and interfaces, then inspect what they receive and return. Now, I have AI write the code and review what it produces. I still decide what to build, check the changes and take responsibility for the result.

My interest goes back further than the current coding tools. By my recollection, I was experimenting with the OpenAI API and Playground roughly two years before ChatGPT arrived. The tools have changed quickly over the last couple of years, and my way of working has changed with them.

ctxforge grew from the context question. I wanted a model to use its engineering knowledge to ask the questions a senior engineer would ask before starting a project: what are we trying to achieve, what is missing and what should stay outside the scope?

In a React and Express prototype, I separated image analysis from product-copy writing so I could inspect each step. With voice agents, I worked on what an assistant should remember when a call passes to another person. The finance review demo starts with a smaller question: can every number in an explanation be traced back to a published statement?

Software that has to keep working

At Knowify, I worked on the frontend of construction software with Angular and TypeScript, and on browser tests with Playwright and GitHub Actions. Features had to fit existing workflows and continue working as the product changed.

At Gadgeon Europe, I worked on Ionic and Angular mobile applications and web dashboards for banking and construction customers. Those dashboards combined several APIs for energy monitoring and connected-device control, including lighting, access and sensor systems. My work also included maintenance and mobile store releases.

Earlier, at Transfutur, I built a transport and logistics ERP from zero as the only software specialist. I worked directly with the owner and office users, then stayed responsible for deployment, operation and support. The system handled roughly €5–6 million in annual invoicing across two offices. I ran and supported it for about a year while the company grew, before it moved to a larger vendor’s system. I am proud that I built something the business could rely on for that stage of its growth.

Business, finance and other interests

My degree is in business and finance. At university, I also volunteered to lead practical informatics classes for a couple of years. That mix still shapes my interests: understanding the work people do, explaining it clearly and finding useful questions to test with software.

I research time-series data and global news around crude oil, and experiment with creative tools. Between Us is a short dialogue-film workflow where voice delivery and camera framing mattered as much as generating the video.

I use AI throughout this work. I need to inspect the code, explain the decisions and check the output before relying on it. The articles include revisions, results and questions I have not answered yet.