Principal AI Systems Architect · São Paulo, Brazil · Remote-friendly
I build AI systems that ship — and remember.
Twelve years building scalable platforms — now focused on applied AI systems architecture: agent orchestration, cognitive memory, RAG pipelines, and generative UI. I take foundation models out of the chat box and into persistent, autonomous products that run in production.
12+
Years shipping software
Platforms, distributed systems, data pipelines
Production
B2G AI platform live
Strata serves city profiles incl. Porto Alegre/RS
2.8M ops/s
State-fork allocation
multiverse-js benchmarked at 430ns per fork
+50pp
Typed memory vs flat retrieval
Blind validation, p < 10⁻⁷ — EPCG Memory Access Model
Selected work
Proof, not promises.
One production B2G platform, two open-source systems, and an embodied AI companion — all attacking the same long-horizon problem.
Hub Esfera · Production B2G
Strata
A municipal intelligence platform with an embedded AI copilot.
- Live in production serving city profiles including Porto Alegre/RS
- Data flywheel: bureaucratic PDFs, gazettes, and spreadsheets into Parquet + agent-optimized Markdown
Open source · Benchmarked
Multiverse ↗
Git-style branching semantics for synchronous UI state.
- Fork allocation at 2.8M ops/s (430ns) — 118,000x mutation cost ratio
- LWW merge optimized to 44.9K ops/s (26µs) — inside a 60fps frame window
Embodied AI · Real-time voice
Pixie
An embodied 3D companion with bidirectional live voice.
- Sub-latency WebSocket voice loop with real tool execution in 3D space
- Audio-amplitude-driven lip-sync, no precomputed visemes
Low-resource NLP · Open
Ayvu-Talian ↗
A language model trained from scratch for Talian.
- Every layer understood — no pretrained shortcuts hiding the constraints
- Tokenization for unstable orthography and heavy Portuguese code-switching
Research · Empirically validated
EPCG
Typed memory interfaces for agents that must remember over months.
- +50pp over flat retrieval in blind validation (N=200, p < 10⁻⁷)
- External AMA-Bench pilots: 72.5% vs BM25 70.0%, LongContext 50.0%
Experience
Twelve years. One direction.
Sep 2024 — Present
São Paulo, Brazil · Hybrid
Co-Founder & Principal AI Architect
Hub Esfera
Architected Strata, a municipal intelligence platform for the public sector (B2G) with an embedded AI copilot — deployed on Cloud Run, serving production city profiles with live data pipelines and agent-driven decision support.
- Designed multi-source public-data ingestion (BigQuery, PNCP, FNDE, INEP) into an embedded analytical DuckDB layer for ultra-low-latency agent queries.
- Built a deterministic LangGraph agent with skill-based reasoning, sandboxed read-only analytical tools, and full LangSmith observability across education, procurement, and budget domains.
- Engineered data-flywheel pipelines converting unstructured bureaucratic datasets (high-volume PDFs, official gazettes, Fundeb spreadsheets) into structured Parquet snapshots and agent-optimized Markdown.
Jan 2023 — Present
Remote
Lead Applied AI Architect
Independent AI Systems Research
Ongoing applied R&D across agent memory, state management, generative UI, and multimodal AI — from experimental architectures to production-ready open-source libraries.
- EPCG Memory Access Model: falsified flat retrieval as a universal memory interface; four typed operations beat it by +50pp in blind cross-domain validation (p < 10⁻⁷), with external AMA-Bench pilots audited end-to-end.
- LoomDB: temporal memory as a formally bounded dissipative system (asymptotic saturation, sub-stochastic propagation) in Rust/WASM; hybrid retrieval policies optimized by evolutionary search instead of hand-tuned constants.
- multiverse-js: Git-style branching state library — fork allocation at 2.8M ops/s, LWW merge optimized from O(N) scan to 44.9K ops/s.
2023 · 5 months
Campinas, Brazil · Remote
Senior Developer Analyst (Contract)
Bosch Digital
Enterprise data integration and infrastructure modernization across complex SAP ecosystems, backend services, and distributed systems.
- Architected impact-analysis protocols for critical data pipelines, presenting technical strategy directly to German headquarters.
2006 — 2015
Brazil
Software Engineering Foundation
SeePix · Martin Luz · EgypTeam · Interactus
Nine years of production engineering before AI: distributed delivery platforms, high-traffic campaign sites, cloud-native Python, and interactive media for global brands.
- SeePix (2014–15): highly distributed digital-signage delivery with edge isolation and robust content sync (PHP/Symfony).
- Martin Luz (2012–13): Tech Lead on high-traffic seasonal campaigns for Walmart and Sony; drove internal Agile and CI/CD adoption.
- EgypTeam (2008–11): cloud-native Python on Google App Engine; foundational data pipelines for early RL environments (Unity3D).
Notes from the builder side.
Notes on memory, context, infrastructure, and the realities of building AI systems outside toy demos.
Essays on memory architectures, context engineering, and what long-running AI systems actually require — arriving soon. Subscribe via RSS to read them first.
Building something that needs to remember?
I work with teams where persistence, context, or long-horizon behavior is the hard architectural problem — and I'm open to principal-level roles with the right team.