Marius F.

Agentic AI&RAG certified | AI/ML

I design and build AI systems where correctness, stability, and failure modes matter more than demos or buzzwords. My work follows a simple rule: if a system cannot be inspected, tested, and reasoned about, it does not belong in production. I have 15+ years of professional experience as a software engineer across financial systems, healthcare and biotech, and security-sensitive environments. That background trained me to think in terms of assumptions, constraints, and measurable behavior. I build full pipelines, not isolated models: data ingestion and preprocessing, structured extraction, classical ML and predictive models, and LLM-augmented reasoning layers integrated into existing data workflows. I work hands-on with modern LLMs, but I treat them as components, not oracles. When a single model is insufficient, I design event-driven services, vector search and RAG pipelines, or custom agent-style logic with explicit boundaries and evaluation criteria. The objective is consistent behavior under real-world conditions, not clever one-off prototypes. Alongside AI systems, I build professional algorithmic trading software. Within established platforms such as NinjaTrader and MultiCharts .NET, I develop advanced indicators and strategies that integrate cleanly with their native backtesting and execution engines. When clients need dedicated or non-platform-bound solutions, I design and implement the full trading stack from scratch. This includes custom backtesters and optimizers, data pipelines, execution logic, and live integration with broker infrastructure such as Interactive Brokers. In both cases, my focus is on robustness in live trading and understanding why a strategy works or fails, not on producing attractive backtests. Where AI genuinely adds value, I integrate it into trading systems in a controlled way: signal aggregation, regime detection, news or sentiment inputs, and decision layers that combine multiple weak signals into something testable. I am deliberately skeptical about unnecessary complexity and careful about where generative models help versus where they introduce noise. I work with a skin-in-the-game mindset. I care about edge cases, operational clarity, and systems that can be maintained by someone other than me. If you need a partner who can turn data, models, or trading ideas into reliable software that ships and runs reliably, I can take your project from first principles to a production-ready implementation.