Todor Evtimov
Bridging raw data and human experience through calm product thinking, useful automation, and carefully designed artificial intelligence.
Selected Work
AI Product System
An AI-powered meal planning and grocery optimization platform designed to help users decide what to cook, what to buy, and how to reduce cost and food waste.
AI generates meal options. Deterministic systems enforce constraints around nutrition, ingredients, and shopping lists.
Applied AI Infrastructure
A lightweight setup for running local language models through LM Studio and exposing them securely for use in custom applications.
Bridges the gap between local model experimentation and practical application infrastructure.
Organizational Tooling
Structured tools, prompts, and workflows that help business teams move from AI experimentation to repeatable, useful adoption.
The aim is not to showcase AI, but to make it usable inside existing business processes.
Enterprise Systems
Data-driven procurement systems supporting reporting, supplier decisions, and operational standardization across multi-country environments.
Built around real-world constraints: fragmented data, multiple stakeholders, and decision-heavy business processes.
ABOUT
I build systems at the intersection of product, data, and applied AI.
My work focuses on turning complex operational environments into structured, scalable, and decision-ready systems. This includes designing data architectures, building internal platforms, and integrating AI into real-world workflows.
Currently, I lead a cross-functional product and engineering setup responsible for procurement platform solutions across multiple countries. Alongside this, I explore how AI can move from experimentation into practical, production-grade tooling.
I'm particularly interested in:
Now Exploring
Fine-tuning compact models to run efficiently while preserving useful reasoning quality.
Developing workflow tools that integrate into real developer and operator routines.
Connecting documentation, context, and knowledge into systems that people can actually query.