← Back

Stephanie Rosales Diaz

NebraskaOpen to relocate (US research & engineering hubs)(260) 267-5541connect@stephanierd.dev

github.com/st3phablestephanierd.dev

Download archived PDF (deprecated)Updated August 2026

The Throughline

The throughline is scientific curiosity about the mind. Dual training in computer science and psychology, production software at Microsoft, and independent systems work at ASTRONOMA all fold into the same question: how do you model belief, inference, and the brain well enough to actually understand them — and build software that holds up under that question? These days that lives in Polaris (a local-first production life system with a questline DAG engine and Plaid-backed finance) carried alongside schizophrenia transcriptomics, hierarchical generative models of pathological inference, and CS / psychology coursework at Purdue.

Research Interests

Computational neuroscience and computational psychiatry: generative / predictive-coding models of belief and psychosis-like inference; neural dynamics; transcriptomic approaches to schizophrenia pathophysiology. The through-question is scientific understanding of mind and brain — reached through modeling, analysis, and the research software that makes it inspectable.

Research Projects

Polaris — local-first production life system (active)

SwiftUI, SwiftData, Node.js, Plaid API · ASTRONOMA

  • Native macOS app that turns long-horizon goals into daily execution: questlines, mental health, household, personal care, deadline map, finance, and a stardust economy.
  • Questline DAG engine (Q2.1–Q2.2): cycle detection, topological order, execution frontier, multi-parent unlock, and a LOCKED → ACTIVE → IN_PROGRESS → COMPLETED / FAILED state machine — pure Swift, independent of the UI tree.
  • SwiftData persistence bridges for tracks and questline blueprints; domain stores for mental health, household, personal care, finance, and economy across a 60+ file Swift codebase.
  • Plaid Link companion (Express): link-token create, public-token exchange, transaction sync with cursor persistence, Keychain rehydration after restart, production OAuth redirect path.

Schizophrenia transcriptomics — differential expression & gene networks (in progress)

Python, SciPy, pandas, GEO GSE21138

  • Framing: diagnostic categories lack mechanistic validity; study dysregulated expression in schizophrenia using public post-mortem prefrontal-cortex transcriptomics (GEO GSE21138; PsychENCODE context).
  • Pipeline: NCBI FTP fetch → forgiving metadata parse → row-wise t-tests (p-value, log2 fold change) → volcano plot; public write-up at github.com/st3phable/schizophrenia-gene-pipeline.

Hierarchical generative model of psychotic-like inference (active)

Python, Bayesian / predictive coding

  • Research question: under what precision regimes does a hierarchical Bayesian agent form stable false beliefs despite sensory evidence?
  • MVP: 2-level generative agent; mis-set sensory vs prior precision → pathological belief trajectories; simulation + write-up (computational psychiatry framing).

Neural activation & E/I-balance visualizer (public demo)

TypeScript, React, dynamical systems

  • Interactive in-silico sketch of excitatory/inhibitory flow (dopaminergic gain vs GABAergic threshold) with homeostasis / hyperactive / hypoactive regimes. Public module on this site; source at github.com/st3phable/neural-visualizer.

Engineering & Work

ASTRONOMA — Founder / Systems Engineer

Remote · 2026 – present

  • Building Polaris as a local-first product: SwiftUI client, Node Plaid backend, SwiftData persistence, and a DAG execution engine with cycle detection and topological unlock.
  • Shipped product shell — Dashboard, Mental Health, Personal Care, Household, Questlines, Shop, Map, Finance — with user-editable CRUD instead of hardcoded presets.

Microsoft — Software Engineer Intern

Azure AD Identity Governance · Remote · Jul 2021 – Aug 2021

  • Integrated a third-party workflow into a production identity platform — large codebases, careful abstraction, reliability.
  • What stuck: reproducible pipelines, careful interfaces, and working inside complex systems — the same habits the research software and Polaris now run on.

Microsoft — EXPLORE Intern

Azure AD Identity Governance · Remote · Jul 2020 – Aug 2020

  • A team software project in enterprise identity — first real exposure to how engineering gets done at scale.

Best Buy — Computing Sales Specialist → Sales Shift Lead

Fort Wayne, IN · Aug 2024 – Mar 2025

  • Earlier: Home Theater / Sony rep, Oct 2023 – Jul 2024.
  • Explaining technical things to real people under pressure, then leading and coaching a floor team.

Education

Purdue University Fort Wayne

Fort Wayne, IN · Computer Science & Psychology coursework (Honors) · 2024–2026

Relevant: Discrete Mathematics, CS II (Java), C & Unix, Psychobiology

Pomona College

Claremont, CA · Computer Science coursework · Apr 2019 – Mar 2022

Relevant: Cognitive Science, Linguistics, Functional Programming (Coq)

LEDA Scholar, Cohort 14

Princeton University · Selective national program for high-achieving first-generation / low-income students. · 2017–present

Skills

Systems & product
Swift, SwiftUI, SwiftData, TypeScript, Node.js, Plaid API, local-first persistence, DAG / topological execution
Scientific computing
Python, NumPy / SciPy / pandas, PyTorch, GEO / RNA-seq pipelines, Git for research
Domains
computational psychiatry, predictive coding / generative models, schizophrenia literature, identity systems (Azure AD)
Also
Unix, Coq / functional programming, Next.js, Bayesian inference (applied), dynamical systems, reproducible research documentation