About
=====

I work the seam where technical products meet the people who need them: the
documentation that teaches them, the positioning that reaches them, and the
analytics that prove what worked. My home turf is **marketing analytics and
machine-learning products** — I write about the kind of software I came up
building and using. My background is an unusual blend of three degrees, and
each left me with something I use every day: a BSc in Economics (Athens
University of Economics and Business), an MSc in Data Science (American
College of Greece; `projects from that degree
<https://c-kapsalis.github.io/ms-data-science-projects/>`_), and an MSc in
Marketing (Syracuse University, as a Fulbright scholar and Beta Gamma Sigma
inductee).

Economics taught me to approach a problem whole, in a structured way. Data
science taught me to read, test and reason about code, and to speak with the
engineers who write it. Marketing taught me to start from the reader: to work
out what someone actually needs before writing a word. Documentation is where
all three finally do the same job.

Domain expertise
----------------

**Marketing and digital-marketing analytics**, my center of gravity. Paid
search and performance marketing (SEM, SEO, email, paid social); marketing
measurement (marketing-mix modeling, customer lifetime value, CAC, ROAS); and
the platforms behind it (Google Analytics 4, HubSpot, Power BI, Looker Studio).
An MSc in Marketing and years working as a marketing analyst and engineer,
documenting analytics for marketers who are not technical.

**Machine learning and data science.** Applied ML in Python (pandas,
scikit-learn); NLP and transformers (BERT, spaCy); retrieval-augmented
generation (LangChain, ChromaDB); and LLM and agent tooling (the Model Context
Protocol). An MSc in Data Science, and hands-on work building an LLM-to-data
integration layer in production.

**Documentation as code.** Git, Sphinx, reStructuredText and CI-published docs,
structured with Diátaxis, the craft that carries the two domains above.

I acquire a domain the way good documentation demands: formal training where I
have it, hands-on use of the product, and structured conversations with the
engineers and specialists who know things they cannot always articulate.

AI fluency
----------

AI is not a side interest for me; it is part of what I build and how I work. At
`Maya Insights <https://mayainsights.com>`_ I helped build a conversational MCP
layer that lets marketers
query a data model in natural language, so I have seen first-hand how completely
the quality of an answer depends on the quality of the documentation beneath it.
I use LLMs throughout my own documentation workflow (drafting, editing, and
pressure-testing structure), and, more concretely, I encode a project's or a
company's documentation guidelines into reusable AI skills that I invoke
iteratively against each draft, so a piece is checked against the house style,
terminology and structure rules before it ships. I treat prompt engineering as
what it really is: a documentation problem. Getting a model to produce consistent, accurate
output takes the same precision, structure and audience-awareness as getting a
product to produce consistent, accurate documentation.

Contact
-------

- **Email:** ch.kapsalis@outlook.com
- **LinkedIn:** https://linkedin.com/in/christoforos-kapsalis
- **GitHub:** https://github.com/c-kapsalis