AI and software engineer: taking ideas through to production — backends, data, model quality, and the tooling around LLMs.
🧪 Off the clock: reproducible math and numerical experiments (Python / JAX).
🖥️ Day job: classic full-stack work — APIs, databases, stitching UIs to services — the boring, reliable kind that keeps production calm.
On top of that:
- 🤖 AI in production — agent-style workflows, integrations with storage & services; latency & cost under control
- 🛡️ ML / AI reliability — tests, guardrails, sanity checks wherever models meet users
- ⚙️ Automation — pipelines, internal tools, demonstrators when they buy time
🛠️ Stack: Python · TypeScript · SQL · Go · Rust
☁️ Cloud: AWS · GCP · 🎨 frontends when needed (React, Vue)
| Focus | Repo | Notes |
|---|---|---|
| 📄 RAG / document ingest | bigos | High-level PDF → markdown for RAG (Docling), OmniDocBench-style evals — early PoC, APIs still fluid |
| ♟️ CGT / Conway verification | complexity-foundations | Algorithms + notebooks + Lean folder; rounds B/C on synergy & topology — active research |
| 🌊 Patterns & JAX | continuous-patterns | Cahn–Hilliard Model C, sweeps, diagnostics (e.g. Liesegang-like spacing) — pre-publication research code |
| 🔗 NL ↔ formal proofs | kalinov-bridge | Lean 4 + LLM loop, optional ForTheL → Naproche path, Gherkin-style tasks — WIP, CI with Mathlib |
| 🎨 GPU viz (Rust + WASM) | phantomat | wgpu rendering, PyO3 / Jupyter, TS packages + WASM — early development |
| 🥒 LLM → DSL (monorepo) | pickled-spec | Shared gate/oracle pattern across BDD, OpenAPI, policy, IaC — pre-alpha |
| 🔢 Exact verification | quantum-foundations | Rigorous Python — sedenions / octonions, generation graphs, CLI (qf-verify-*), papers/ |
| 🧪 Supramolecular + QC helpers | suprakit | xtb/tblite, geometry, Typer CLI, MkDocs; long-term NL → chemistry DSL roadmap — alpha |
💡 TL;DR: nonlinear phase-field PDEs + polymorphic pinning in reaction–diffusion — pattern formation × geological soft matter, spectral GPU codes, plus ML / AI where it earns its keep (simulation, inverse problems, synthesis).
🧩 Also: theory of computation & Turing machines, consciousness (what it is / might be), and conceptual foundations of AI — not only bigger models.
🔭 Physics
- 🌀 Self-organization & patterns (Cross–Hohenberg)
- 🪨 Geological soft matter (Cartwright / Sainz-Díaz)
- ⚡ Far-from-equilibrium thermodynamics & dissipative structures
- 🔀 Phase transitions — order–disorder & morphological (agate → labyrinth, γ)
- 💧 Reactive transport — diffusion–reaction–precipitation in porous media · LBM (stage‑2 coupling)
📐 PDEs · numerics · inverse
- 📊 4th-order nonlinear PDEs — Cahn–Hilliard, multi–order-parameter (H / B, Hohenberg–Halperin)
- 🔬 Reaction–diffusion — Turing / Liesegang vs Jabłczyński-type spacing breaks
- ⚖️ Calculus of variations — free-energy + polymorphic pinning
- 📡 Spectral: FFT + ⅔ dealiasing · IMEX · mass control
- 🌀 Bifurcation / linear stability — precipitation fronts → labyrinth routes
- 🚀 GPU + JAX (autodiff) toward inverse fits from band-spacing data
♟️ Computation theory · TCS
- Theory of computation & Turing machines — Church–Turing narrative, classical complexity (time / space, reductions)
- P vs NP · combinatorial game theory · BSS & surreal machines · Herlihy–Shavit / game complexes (topological complexity)
🧠 Consciousness · mind · AI foundations
- Consciousness — what it is / could be: philosophy of mind, mechanistic & neural angles, and where explanations still fall short
- Foundations of AI — learning & representation beyond “train a bigger net”: generalization, agency, definitions of intelligence, ties back to computation theory
✅ Methodology
- 📜 Lean 4 formalization · 🔢 exact-arithmetic verification · 📉 LOO-CV & null-model tests
🔗 Algebra & foundations
- 🔷 Sedenions · octonions · Jordan J₃(O) · Clifford · Fano geometry
🚶 Walks · ⛰️ mountains · 📚 books — staying curious about science & tooling.
📫 Contact: contact@bartrosa.dev · 🌐 bartrosa.dev · ORCID · Zenodo