About
Software engineer, mathematician, and builder of trustworthy AI systems—credentials behind Artificial Quirks.
I help teams ship agent systems and production GenAI with the same architectural rigor, testable invariants, and runtime checks you would expect of any enterprise platform—not demo stacks that collapse under cost, evaluation, or audit pressure.
The writing on Artificial Quirks is the publication. Services covers optional, scoped advisory when you want that lens on your stack.
I began in software engineering over two decades ago, grounded in a Bachelor’s degree in Mathematics. The path has tracked major industry shifts—from desktop applications and N-tier .NET systems to cloud-native TypeScript microservices and, most recently, agentic AI and deep learning systems.
Technical Journey#
Enterprise Architecture & Conversational Platforms#
As a Lead AI Solution Engineer and Architect, I spearheaded the modernization of conversational platforms toward Generative AI and LLM-based designs. This involved authoring agentic frameworks, defining risk vectors, and leading the transition from traditional microservices to AI-driven customer service solutions.
Related: platform & risk framing on Services, and the ongoing RLM series on runtime policy under real budgets.
AI Startups & RAG#
As a co-founder of Smart City BG, I engineered RAG-based AI assistants designed to support local municipality administration. This gave hands-on experience building local, private LLM pipelines using Ollama, LangChain, LangGraph, and vector database structures.
Quantitative Research & Applied Mathematics#
A mathematics background (B.Sc.) has guided work toward quantitative analysis and algorithmic trading: systematic strategies with deep reinforcement learning, temporal convolutional networks, and statistical tools such as Kalman filters and Bayesian inference. That same systems-thinking habit shows up when agent telemetry is messy and proxy metrics mislead.
Core Expertise#
- AI & Machine Learning: LangChain, LangGraph, Retrieval-Augmented Generation (RAG), ChromaDB, Ollama, Model Post-training, PyTorch, Kalman Filters.
- Languages & Frameworks: Python, TypeScript, Node.js, React, C# (.NET), SQL, MongoDB.
- Methodologies: System Dynamics, Discrete Optimization, Algorithmic Analysis, Microservices Architecture, CI/CD (Docker, Kubernetes).
Lifelong Learning & Certifications#
Recent focus areas include:
- State Estimation & Machine Learning: Kalman Filters (Linear, Nonlinear, and Parameter Estimation) — University of Colorado
- Generative AI & LLMOps: Post-training LLMs, Building with Haystack, Automated Testing for LLMOps — DeepLearning.AI
- Quantitative Analysis: Python and Machine Learning for Asset Management, Advanced Portfolio Construction — EDHEC Business School / Columbia University
How to work together#
For architecture reviews and advisory work, start from Services. Send a short written brief: system under review, failure mode, success criteria, and hard constraints (cost, compliance, timeline). Engagements are scoped—not open-ended staff replacement.
Industry and agent-system notes live on the home page and in the blog. Claims stay labeled; numbers come with horizons and limits.