Silvia Onofrei Lab

Silvia Onofrei, PhD

Applied AI through a mathematician's lens

About

I'm an applied AI scientist, mathematician, and founder of Silvia Onofrei Lab.

After a decade in academia, I architected and deployed AI systems across healthcare and finance at John Snow Labs, SAE, and Qarik (now Egen).

I built clinical knowledge graphs, fine-tuned LLMs for text-to-Cypher generation, and designed custom transformer architectures for tabular data. I also engineered hybrid RAG pipelines for enterprise search and retrieval, and agentic systems for complex data analysis and reasoning.

Writing is part of how I develop and refine my thinking. That is why I started Matheinated, where I cover NLP, graphs, mathematical approaches to machine learning, and the projects I am currently exploring.

Outside of work, I have a deep interest in neuroscience-informed horsemanship. I live in Colorado with my husband and our two horses, enjoying life on the edge of the wilderness.

Silvia Onofrei

Focus Areas

NLP & LLM Engineering

Information extraction, custom transformer architectures for text and tabular data, and post-training through supervised fine-tuning, preference optimization, and domain adaptation.

Graph ML & Knowledge Graphs

Ontology and schema design, spectral methods and advanced graph algorithms, graph representation learning.

GenAI Systems

Hybrid retrieval (Vector + GraphRAG), agentic workflows, deployment, evaluation, and failure analysis.

Silvia Onofrei Lab

Independent judgment for complex AI systems

An applied AI research practice for
organizations facing high-stakes
AI decisions in regulated industries.

About

Most AI initiatives fail from poor framing, fragile architecture, or shallow evaluation. Model choice is rarely the problem.

The Lab helps your team determine what to build, how to structure it, where risk will surface, and what not to build before the roadmap locks you into the wrong solution.

The Lab specializes in language models, knowledge graphs, and the retrieval and agentic systems built on them. It is open-source forward and engineered to fit the stack you already have.

Engagements

Advise

Technical advisory for leaders evaluating AI adoption, surfacing hidden risks, over-engineering, and stack decisions before capital is committed.

Evaluate

Custom benchmarks, observability, and hallucination reduction that expose the edge-case failures and non-deterministic behavior a demo masks.

Discover

Research-to-production translation for problems with no off-the-shelf answer, from feasibility testing to custom prototypes built against real engineering constraints.

Enable

AI education for executive, technical, and non-technical teams, building the foundational understanding and realistic technical judgment that AI decisions require.

Contact

Silvia Onofrei Lab, LLC
Boulder, Colorado

{{ emailText }}

Socials