Hello Stranger!
Welcome to my corner of the Universe.
A place where starlight meets code, and where silence meets signal.

Portrait of Olga Moreira

Astrophysicist • AI Architect • Data Analytics & Modular Systems

Scalable Insights from Complex Data
Open Science Advocate

Publications

About Me

Hi! I'm Olga Moreira, an astrophysicist and AI knowledge architect.

I design decentralized, portable, modular systems. My work bridges astrophysics with semantic systems for machine-learning knowledge bases.

I hold a PhD + MSc in Astrophysicsfrom the University of Liege and a BSc + MSc in Physics & Applied Mathematics (Astronomy) from the University of Porto. I've worked with both the European Space Agency (ESA) and the European Southern Observatory (ESO). Based in the Niagara region, I now focus on instruction-driven, auditable AI workflows and science-communication projects that unite logic, openness, diversity, and inclusivity.

My philosophy is rooted in local-first, eco-conscious design. Systems that run on user devices, adapt to low-resource settings, and reduce dependence on energy-intensive centralized infrastructure. For me, decentralization and modularity are more than engineering choices: they're a path towards accessibility, portability, privacy, and a lighter environmental footprint.

Consulting & Collaboration

Strategic systems for data-heavy and AI-driven teams.

I help teams design reliable, interpretable, and well-structured AI and data workflows.

My work blends astrophysics-level analytical rigor with modern AI systems architecture, giving clients clarity, reproducibility, and measurable results.

How I Work

  • Clear scoping and expectations
  • Transparent, reproducible methods
  • Clean documentation
  • Honest, actionable feedback
  • Collaborative, detail-oriented approach

My goal is to deliver systems that are robust, easy to maintain, and genuinely useful.

Data Modelling & AI Integration

AI-Assisted Editorial & Knowledge Systems

I design structured workflows that make technical content accurate, consistent, and scalable.

  • Modular prompt and instruction design
  • Semantic taxonomy alignment
  • Quality checks and audit trails
  • Local-first or hybrid LLM integration (Ollama, HF, OpenAI-compatible)

Data Analytics & Decision Modeling

I turn complex datasets into clear, defensible insights using transparent, well-documented methods.

  • EMV/EVPI decision analysis
  • Bayesian reasoning (e.g., PPV, diagnostic metrics)
  • Regression and classical ML models with diagnostics
  • Reproducible Python pipelines and clear reporting

Modular AI Assistant Architecture

I design lightweight, privacy-conscious AI assistants tailored to real workflows.

  • Chat and multi-step agent modes
  • Lightweight databases (SQLite)
  • Reusable modules and consistent reasoning patterns
  • Local + cloud LLM routing

Ideal for teams seeking portable, transparent AI assistants that enhance productivity without compromising data privacy.

Cognitive AI Systems & Contextual Memory Architecture

Assistant Brain Architecture

A lightweight, modular reasoning engine for LLM-based assistants that integrates:

  • Instruction and reasoning layers
  • Error-taxonomy checks
  • Session continuity via lightweight memory stores
  • Structured self-reflection and correction

This architecture underpins my GM-RKB workflows and my personal assistant system ("Sam AI"), supporting transparent, auditable AI-human collaboration.

Neuro-Tag Cognitive System

A structured tagging framework that encodes:

  • Emotional-state classification
  • Boundary and trigger detection
  • Truth-checking logic
  • Semantic cues for reasoning and safety
  • Contextual recall rules

Originally developed for trauma-aware AI interaction, Neuro-Tags now serve as a reusable cognitive layer for agentic LLM systems.

Start a Collaboration

If you have a project in AI, data analytics, or knowledge-system design, I'd be happy to review it.

Send me your project idea, I'll prepare a proposal and budget estimate.

Let's Scope Your Project

Projects / Portfolio

A few of the systems and analyses I've been building at the intersection of starlight, data, and AI.

GM-RKB Modular AI Editorial Assistant

Local-first AI workflows for machine learning concept pages.

  • Designed modular prompt and instruction layers for 500+ ML concept pages.
  • Built a SQLite tracker to monitor concept completeness and editorial priorities.
  • Integrated local and cloud LLMs (Ollama, HF Transformers, OpenAI-compatible APIs).
  • Implemented semantic quality checks for taxonomy, links, and references.

Decision Analysis & Regression Modeling

EMV/EVPI, Bayesian PPV, and regression analytics in Python.

  • Developed EMV/EVPI decision models with salvage-value and probability sensitivity analysis.
  • Implemented Bayesian PPV calculations for diagnostic testing scenarios.
  • Built OLS and Ridge regression models with R², RMSE, and residual diagnostics.
  • Delivered clean CSV tables, PNG plots, and fully vectorized Python scripts.

Sam AI — Modular Personal Assistant

Local-first personal assistant with structured reasoning and portable memory.

  • Combines chat mode with agent mode for multi-step workflows.
  • Uses a portable contextual memory layer that can run locally or via APIs.
  • Includes truth-checking and cognitive-distortion detection for safer reasoning support.
  • Designed as a modular system so new tools and workflows can be added easily.

Solar Physics Analytics Notebook (In Progress)

Integrated analysis of stellar, flare, and sunspot datasets.

  • Combines SIMBAD stellar parameters, GOES solar X-ray, and NOAA sunspot region data.
  • Plans regression modeling of stellar and solar parameters with interpretable diagnostics.
  • Explores flare classification metrics (PPV, sensitivity, specificity) and event statistics.
  • Frames results in a decision-theoretic context for space-weather risk assessment.

Open Research & Initiatives

Research threads and long-term projects exploring the intersection of astrophysics, AI systems, and open knowledge.

OM OpenSkies Initiative

A long-term project creating emotionally supportive, culturally inclusive scientific resources. It blends open-access research, Indigenous sky traditions, and modern astrophysics to build a global platform where science is a shared right — not a privilege. This initiative includes future projects such as the "OM Atlas of the Skies", "Voices in Science", and the "SkyMemory Archive".

Solar Physics & Stellar Research

My scientific roots are in asteroseismology, solar-like oscillations in red giants, and interferometric radius calibration — work that connected stellar interiors with observable seismic signatures. I continue exploring these themes while expanding into heliophysics and space-weather research.

I am currently drafting a review paper, From SOHO to Solar Orbiter: 25 Years of Solar MHD, which traces the evolution of solar missions, MHD simulations, and the transition toward data-driven heliophysics. The review aims to bridge classical plasma modeling with modern machine learning workflows and next-generation mission data.

Additional ongoing threads include flare statistics, sunspot region analysis, and the broader interplay between magnetic activity in the Sun and other solar-type stars. These projects evolve steadily alongside my AI systems work, informing how I combine astrophysical reasoning with modern computational tools.

AI Cognitive Systems & Assistant Architecture

Ongoing conceptual development of AI systems that combine:

  • structured reasoning templates,
  • contextual memory engines,
  • cognitive tagging systems (Neuro-Tags),
  • and modular assistant workflows.

This includes the evolving "Assistant Brain" architecture, designed for local-first, transparent, and auditable AI collaboration.

These research threads evolve slowly and steadily alongside my professional work, informing how I design AI systems, analyze data, and explore the sky.

Clear and open skies

I build humane systems at the edge of science and AI. If you're curious, let's talk.