ABOUT
AI Engineering. Software foundation. Product lens.
I’m Marcelo Taparelli, an AI Engineer at Agência Catus and a Software & AI Engineer at EVAG. I work on AI research, process automation, and agentic development, with experience in backend, APIs, production systems, and product. I start with the business problem before choosing the technology.
Verification, security, and operability guide the path to production. I surround probabilistic components with typed contracts, deterministic policies, tests, evals, fallbacks, and human approval. Observability and measurement support improvements within the limits of each evaluation. Product is my lens for deciding why to build, who it serves, which value matters, and which trade-offs to accept.
ENGINEERING METHOD
How I build AI systems.
From business problem to production: domain, architecture, and constraints define where AI fits. Verification, security, and measurement make the system reliable and able to evolve.
- BUSINESS PROBLEM
- DOMAIN MODEL
- ARCHITECTURE
- CONSTRAINTS
- AI / AGENT
ORCHESTRATION - VERIFICATION
- SECURITY
- PRODUCTION
- MEASUREMENT
- IMPROVEMENT
Each case shows how far this method was exercised, with explicit evidence and limitations.
A path of connections
At EVAG, I started as a Product Owner in April 2023, connecting clients, design, and development to deliver websites. In April 2024, I moved into systems development and automation, including features and production fixes. At Agência Catus, I worked as a Full Stack Developer from June to September 2026 on e-commerce and WordPress. Since October 2026, I hold AI Engineering and Software & AI Engineering roles at Catus and EVAG respectively, researching AI-assisted methods and developing automation and systems.
Professional experience
Oct 2026 — Present
Agência CatusAI Engineer
AI research for process automation and product delivery. Agentic development of stores integrated with e-commerce platforms and WordPress sites, including themes, features, deployment, testing, validation, and documentation.
Oct 2026 — Present
EVAGSoftware & AI Engineer
Research, analysis, and development of AI-powered automation to support operations. Agentic development and system maintenance, following software best practices, testing, and security.
Jun 2026 — Sep 2026
Agência CatusFull Stack Developer
Development and maintenance of e-commerce platforms and WordPress sites.
Apr 2024 — Sep 2026
EVAGDeveloper
Petition automation via Google Drive/WordPress and request management integrated with GitLab. Systems with Laravel, Python, JavaScript, WordPress, and Tailwind, including features and production fixes.
Apr 2023 — Mar 2024
EVAGProduct Owner
Website delivery from discovery to deployment, connecting clients, design, and development. Requirements, briefs, workflows, and prioritization across business, UX, and technical capacity.
Capabilities in context
AI Engineering
AI research for process automation and agentic development, with testing, validation, and documentation. Python, FastAPI, scikit-learn, pandas, Machine Learning, LLMs, RAG, embeddings, pgvector, and hybrid retrieval. Agents with LangGraph and OpenAI, structured outputs, guardrails, human approval, and OpenTelemetry observability. Evals with deterministic baselines and frozen held-outs; domain adaptation/fine-tuning with PyTorch, GPU/CUDA, mixed precision, validation-based checkpoint selection, and confidence/calibration analysis.
Software Engineering
Experience with systems, APIs, automation, e-commerce, and WordPress. TypeScript, Bun, Node.js, Laravel, Python, JavaScript, Tailwind, REST APIs, SQL/PostgreSQL, Redis, and Git. Domain modeling, software architecture, Clean Architecture, DDD principles, SOLID, Clean Code, and TDD. Unit, integration, and E2E tests; idempotency, retries, and circuit breakers for concurrency and failures.
Production, security & reliability
OpenTelemetry observability, metrics, authentication, authorization, secret management, resource limits, and graceful degradation. Docker, Terraform, GitHub Actions CI/CD, and artifact-based delivery. Hands-on temporary non-production AWS lab: ECS/Fargate, ALB, RDS, ElastiCache, ECR, Secrets Manager, IAM, and CloudWatch, with migrations, troubleshooting, and post-destroy auditing. Professional work on performance and production software issues; AI cases define their experimental validation scope.
Product perspective
Frame the business problem, understand who I build for, and decide which value to measure. Product Owner experience delivering websites from discovery to deployment, connecting clients, design, and development. Briefs, requirements, workflows, and prioritization across business, UX, and technical capacity. Make constraints and trade-offs in reliability, security, performance, cost, latency, and operability explicit; decide whether AI is necessary before adding it.
Engineering in practice
In OpsPilot AI, I built RAG and agentic workflows with Python/FastAPI, pgvector, LangGraph, and OpenAI, with human approval and OpenTelemetry. In Ops Triage AI, I combined a deterministic baseline, a local LLM, and a hybrid policy; I evaluated Jev and adapted Laya with PyTorch/CUDA in experiments separate from the runtime. In DefectRisk, I worked on leakage-safe ML evaluation, risk ranking, calibration, and uncertainty. In Resilient Transaction API, I worked on idempotency, resilience, and observable operations with TypeScript/Bun, PostgreSQL, and Redis, including a temporary non-production AWS lab.
Education
Postgraduate Program in AI Engineering
In progress · expected completion: April 2027
Systems Analysis and Development
Completed in 2026
Software Engineering — Alura
Back-end · Node.js APIs · Authentication · Testing · Security · DevOps/CI/CD · Cloud/AWS · Secure Development
LET’S TALK
Good products start with a good conversation.
Professional opportunities, partnerships, or a conversation about AI Engineering, software, and product.
contato@marcelotaparelli.com.brConnect on LinkedIn