Cassio – Python, AI agent development, AWS

It's an AI automation platform for healthcare operations.

• Designed and shipped AI-powered workflow automations for healthcare operations.
• Built back-end services and integrations connecting LLM agents, APIs, and customer workflows.
• Collaborated with clients, product, and engineering to define requirements and deliver production solutions.
• Improved operational efficiency by automating repetitive manual healthcare processes.

Open-source AI-assisted crypto trading platform that combines deterministic trading automation with optional CrewAI-powered advisory agents. I designed and built the system to analyze market data, compute technical indicators, classify market regimes, and manage simulated or live trading decisions through an auditable pipeline.

The project focuses on using AI where it adds judgment and context, while keeping core execution deterministic, traceable, and safe. The system supports agents that can evaluate uncertain market conditions, reason about trading opportunities, and provide structured recommendations, while the main trading flow remains controlled by explicit rules, validation layers, and order lifecycle management.

• Designed and implemented the platform architecture in Python, including data ingestion, market analysis, strategy execution, and order lifecycle management.
• Integrated CCXT to retrieve ticker and OHLCV data from multiple crypto exchanges, enabling exchange-agnostic market analysis.
• Built indicator pipelines for EMA, RSI, Bollinger Bands, MACD, and market regime detection, allowing the system to classify conditions such as trending, ranging, volatile, or unknown markets.
• Implemented a deterministic execution layer for paper/live trading modes, including order states such as pending, open, partially filled, filled, canceled, and rejected.
• Added CrewAI-based advisory agents to support decision-making in ambiguous scenarios, separating AI reasoning from critical execution logic to improve auditability and reduce risk.
• Created configuration-driven strategy behavior, allowing the system to adapt profit targets, fees, precision rules, and trading parameters without hardcoding business logic.
• Developed a dashboard/UI to inspect trading status, balances, recent orders, agent outputs, and operational state.
• Focused on practical engineering concerns such as traceability, modularity, fee accuracy, restart resilience, and clear boundaries between automation, AI assistance, and deterministic execution.

It's an event-driven exception management platform for alternative investment operations. The system replaces slow manual triage workflows with Kafka-based event processing, dynamic rule queues, and analyst-facing prioritization tools.

• Built FastAPI services and Kafka pipelines for scalable exception processing.
• Designed dynamic rule-based queues to prioritize and route operational exceptions.
• Integrated AWS Lambda consumers with PostgreSQL persistence.
• Added Datadog observability and Terraform infrastructure automation.
• Reduced average exception handling time from 15 minutes to 90 seconds.

It's a Databricks ETL optimization for enterprise data pipelines (a multinational fast-food holding company).

• Optimized Databricks ETL jobs, reducing runtime by more than 70%.
• Rebuilt and tested AWS Glue jobs to improve reliability.
• Improved Spark-based data workflows for scalability and maintainability.
• Supported Python and scikit-learn data processing workflows.

It's a personalized learning platform serving language learners across multiple modalities.

• Led a small engineering team delivering personalized learning features.
• Built back-end APIs and workflows using FastAPI and PostgreSQL.
• Collaborated with product stakeholders to translate learning goals into technical solutions.
• Improved AWS infrastructure and CI/CD processes, reducing operational costs.

It's a research infrastructure data platform used by universities and research institutions.

• Migrated legacy Pentaho ETL workflows to Apache Airflow.
• Built and maintained data pipelines for academic and research infrastructure.
• Improved search and indexing workflows using Elasticsearch.
• Added monitoring and observability with Grafana and MongoDB-backed systems.