# Work — Caio Theodoro > Employment history, certifications, and open-source and research projects. ## Experience ### Senior AI Engineer — Adopt AI Full-time · Mar 2026 – Present (5 mos) · California, United States · Remote Production AI workflow infrastructure: agent execution, human review, evaluation, and orchestration. - **Agent Execution Harness**: Architected the core runtime — turn and message lifecycle, idempotent claim semantics, and replay-safe Temporal orchestration. Every agent capability the platform ships (HITL, policies, evals, generative UI) runs on it. - **Live Evaluation Dashboard**: Built production conversation-quality scoring that surfaces degradation early enough to intervene before customers feel it. - **Human-in-the-Loop Interventions**: Shipped the escalation system end to end — a signal-driven resolve state machine with optimistic locking and tenant isolation, a 3-action resolution model, and the Temporal signal contract wiring escalations into live agent workflows. - **Generative UI Architecture**: Designed a registry of approved components the harness selects from instead of generating markup each turn, cutting dashboard prompts from roughly 15k tokens to 700–2k; a JSON-Patch delta protocol saved 90% on multi-turn follow-up edits. - **Policy Engine & Multi-Model Inference**: Built agent guardrails with policy-version pinning and full audit logging, and integrated AWS Bedrock behind one selection interface, cutting latency and cost through provider routing and prompt caching. ### Advisor — Avenza Self-employed · Mar 2026 – Present (5 mos) · Remote Advising on technical direction, architecture, and client delivery oversight for the studio I founded. - **Technical Direction**: Advise on architecture, technology choices, and engineering standards for the studio I founded. - **Delivery Oversight**: Review client engagements and solution designs without day-to-day execution ownership. ### Senior Software Engineer — Acorns Full-time · Jul 2025 – Mar 2026 (8 mos) · California, United States · Remote Servicing platform modernization at a consumer fintech serving millions of users, during the move from a Rails monolith to Node services, React frontends, and a GraphQL gateway. - **Full-Stack Delivery**: Shipped production changes across Rails, Node services, React, Next.js, and GraphQL APIs throughout a phased monolith-to-microservices migration. - **Internal Tooling**: Built GraphQL-driven admin and support tooling giving operations teams direct visibility into servicing workflows, and resolved gateway schema and data-consistency failures introduced by splitting Rails-owned data across independent services. - **Ephemeral Environments**: Shipped per-PR environments to production, closing the staging/production parity gap that had been a recurring source of regressions. - **Regression Coverage**: Partnered with platform and QA to extend automated regression coverage across the new Node services, reducing cutover risk during the migration. ### Founder & Senior AI Engineer Consultant — Avenza Self-employed · Feb 2025 – Mar 2026 (1 yr 1 mo) · Remote Founded and led an AI consulting and product studio, run in parallel with full-time roles: agents, computer vision systems, analytics platforms, and automation infrastructure delivered to outside clients. - **Team & Technical Direction**: Directed a 7-engineer team across client discovery, solution architecture, delivery, and engineering standards. - **Production AI Agents**: Architected agents across logistics, customer support, operational workflows, and computer vision analysis, built on structured outputs, retrieval, model-backed classification, and human review loops rather than raw prompt calls. - **Foreign-Exchange Automation**: Delivered end-to-end software and automation for a foreign-exchange house covering operational workflows, back-office processing, and third-party integrations. - **Analytics & Automation Platforms**: Designed platforms spanning data ingestion, dashboards, paid-traffic systems, and business reporting. - **Revenue Impact**: Delivered systems tied to roughly $2.8M USD in client revenue over 8 months. ### Software Engineer — MB Labs Full-time · Aug 2022 – Jul 2025 (2 yrs 11 mos) · Remote Cloud-native products, data systems, and architecture consulting across fintech, education, content, and supply-chain clients. - **Cloud Infrastructure & Performance**: Built production systems on AWS (ECS, RDS, ElastiCache, S3, CloudFront, ALB) with Cloudflare and Redis for low-latency fintech workloads serving millions of users. - **Microservices & High-Availability**: Designed microservices and APIs for high-traffic clients; connection pooling and regional caching cut latency, then right-sizing compute brought infrastructure spend down. - **Event-Driven Data Pipelines**: Architected SQS and async-worker pipelines for compliance-grade fintech operations. - **Key Clients**: GS1, ANBIMA, YOU, Engagx, Evolucional, Panorama. ### Analyst / Software Engineer — Mactus Informatica Full-time · Jun 2021 – Jul 2022 (1 yr 1 mo) · Campo Mourão, Paraná, Brazil Built business systems, internal tools, and modernization work for operational software. - **Legacy System Modernization**: Modernized legacy applications and improved stability for critical workflows. - **Cloud-Ready Architecture**: Designed cloud-ready services with clearer performance, maintenance, and cost boundaries. ### Intern / Full-Stack Developer — ATLA Ensino Internship · Dec 2020 – Jun 2021 (6 mos) · Campo Mourão, Paraná, Brazil Built data extraction, API, and back-office systems for education operations. - **Data Extraction & Processing**: Built Python scrapers that processed 5,000+ records per day. - **API Development**: Developed REST APIs for internal data workflows. ### Fullstack Developer — Haken — Empresa Júnior Contract · Jul 2019 – Jan 2021 (1 yr 6 mos) · Campo Mourão, Paraná, Brazil Delivered web and mobile software for early clients and university-backed projects. - **Project Management**: Managed project scope, delivery plans, and client communication. - **Custom Development**: Built custom CMS and automation features for small-business workflows. ## Projects - **ReconForge** _(AI, Evals, Fine-tuning)_: Laptop-trained 1.7B LoRA for financial reconciliation exceptions — 0.913 severity-weighted recall against a frontier model's 0.872, 1.000 recall on high-severity cases, zero API cost. [source](https://github.com/caiotheodoro/reconforge) - **Suture** _(AI, Vision, Fine-tuning)_: 8B VLM that diffs an underwriting binder against the issued policy: 0.959 recall across 13 discrepancy classes where zero-shot GPT-5.6 vision scores 0.373. [source](https://github.com/caiotheodoro/suture) - **LossBench** _(AI, Evals, Backend)_: Scores money-touching agents on expected operational loss instead of accuracy — severity-weighted, calibrated, replayable, on a hash-chained decision ledger. [source](https://github.com/caiotheodoro/lossbench) - **Habeas** _(AI, Fine-tuning)_: Form I-9 compliance validation fine-tune that cites the violated clause in M-274 / 8 CFR 274a.2, trained on synthetic-only data. [source](https://github.com/caiotheodoro/habeas) - **Substrate** _(AI, Evals, Research)_: Five agent-systems research units, each a reference implementation paired with the benchmark harness that can falsify its own claim. [source](https://github.com/caiotheodoro/substrate) - **Specula** _(AI, Vision, Fine-tuning)_: FDA food-label compliance review: label photo in, cited PASS/FLAG report out, grounded in 21 CFR 101 and openFDA enforcement data. [source](https://github.com/caiotheodoro/specula) - **Plumb** _(AI, Fine-tuning)_: AIA G702/G703 pay-application audit that catches math errors, retainage drift, and double-counted change orders, returning corrected figures per line. [source](https://github.com/caiotheodoro/plumb) - **Perfectman** _(AI, Simulation, TypeScript)_: Social simulation where AI personas notice unevenly, reply late, lurk, and form private alliances — presence over tick loops. [source](https://github.com/caiotheodoro/perfectman) - **Prompt Sharpener** _(Rust, Desktop, AI)_: Tauri overlay on a global shortcut that rewrites a rough prompt into task, scope, constraints, and success criteria — Anthropic, OpenAI, Gemini, or local Ollama. [source](https://github.com/caiotheodoro/prompt-sharpener) - **Readapt** _(Frontend, ML)_: Reading app that adapts text layout for dyslexic users, estimating accessibility needs from a photo. [source](https://github.com/caiotheodoro/readapt) - **Synkro** _(Backend, DevOps)_: Real-time data synchronization middleware on gRPC and Kafka for high-throughput distributed systems. [source](https://github.com/caiotheodoro/synkro) - **Task Circle** _(Frontend, Backend)_: Team task management with gamification mechanics to boost productivity and engagement. [source](https://github.com/caiotheodoro/taskcircle) ## Certifications - AWS Cloud Technical Essentials — AWS, credential KMKFS6VQVVTH - IBM Machine Learning Specialization — IBM, credential ROVWV28S6RHK - IBM DevOps and Software Engineering — IBM, credential KAZJXRVEMYAS - AWS Cloud Practitioner Fundamentals — AWS, credential WH292DLLN5N4 - Google Data Analytics Certificate — Google - AWS Cloud Solutions Architect — AWS, credential 7YW27R22BX2M - Full Stack Software Developer — IBM, credential MZTYTKMZSSTJ ## Elsewhere on this site - [Home](https://caio.theodoro.dev/index.md): Who Caio is, what he works on now, and the current role. - [About](https://caio.theodoro.dev/about.md): Background, how he works, and what he is currently focused on. - [Contact](https://caio.theodoro.dev/contact.md): Email, profiles, what to reach out about and expected response time. - [Privacy](https://caio.theodoro.dev/privacy.md): What this site collects (almost nothing) and which third parties see a request. - [Blog index](https://caio.theodoro.dev/blog.md): Every published post with date, tags and summary. - [Journey](https://caio.theodoro.dev/journey.md): Year-by-year timeline of the path into ML engineering. HTML version of this page: https://caio.theodoro.dev/projects. Machine-readable index: https://caio.theodoro.dev/llms.txt. Full URL list: https://caio.theodoro.dev/sitemap-index.xml.