Lucas Aguiar

Building AI-powered systems for agricultural regulatory compliance. Designing and shipping LLM-driven workflows, agentic experiences, and document understanding pipelines that turn regulatory expertise into software. Stack: FastAPI, Temporal.io, Langfuse, LLMs (Claude, ChatGPT), RAG on PostgreSQL.

Core member of the Reports / Data Visualization team, responsible for analytics infrastructure serving 2 billion time entries across Toggl's product suite.

• ETL & Analytics Pipeline

  Before this system, large users couldn't even pull monthly data — most requests just timed out. Designed and shipped an ETL pipeline that made "all-time" reports possible for the first time. Worst-case response dropped to under 3 seconds.

• DSL Report Engine

  A configuration-driven engine where adding a new report means writing a handler and a service — nothing else changes. Three products (Track, Work, Focus) run on it today, fully decoupled. Onboarding a second product took days, not months.

• Historical Billable Rates

  Temporal billing across 4 hierarchy levels (Workspace → Workspace User → Project → Project User) with flexible effective dates. ~200M time entries affected. Users pay employees based on this data, so there was zero margin for error.

• Calendar Integration

  Sync with Google and Outlook Calendar — calendar events show up as time entries.

• Reports V2 → V3 Migration

  Replaced the legacy reports infrastructure with v3 internals while keeping every endpoint alive. Clients saw zero breaking changes thanks to an internal abstraction layer.

Building products from zero to production — architecture, development, and infrastructure. Focus on creating self-sustaining systems that require minimal maintenance.

• Vodsafe — Video On Demand Platform

  Distributed video streaming platform for security monitoring: live streaming, historical playback, multi-camera management. Currently serving 84+ cameras. Master server (Django) + scalable Stream servers (Go + OvenMediaEngine). Frontend built with React and Tailwind CSS. Self-provisioned bare-metal infrastructure via Ansible.

• Ágrafo — Legal Document Automation

  Document generation platform for law firms. Built template engine that compiles legal documents from configurable models. Frontend built with React and Tailwind CSS. Reduced document preparation time by 78% (90 min → 25 min). Zero maintenance required since launch.

Migrated legacy PHP ETL system to Python, improving maintainability and performance.

Led technical and managerial decisions for Processamento, an ETL system processing official court orders from Brazilian tribunals. Implemented a Deep Learning solution, a Convolutional Neural Network (TensorFlow) modified specifically for this purpose, to automatically extract legal publications from official gazettes. Replaced a team of 20+ manual processors, recouping 8 years of team costs within 6 months. Focused on team development, delivery consistency, and process stabilization.

Built Andamentos, a Python web crawler extracting lawsuit data from all Brazilian courthouses — millions of requests daily. Scaled through database indexing, profiling, Docker optimization, and proxy management. Later initiated Processamento, a microservices-based ETL system for court PDF processing.

Built an API from scratch with automated infrastructure provisioning (Ansible), full test coverage, and a web crawler for large-scale database scraping using SOCKS5 proxies.

Fully responsible for maintaining a SCADA system, solving non-conformities and other legal tasks, dealing with instrumentation and automation. The system maintained was responsible for releasing the product batch documentation.

Worked with validated systems, instrumentation, and industrial automation in a GMP environment.