Caio is available for hire
Hire CaioCaio Vitor Oliveira Moreira
Verified Expert in Engineering
Data and AI Engineer and Developer
São Paulo, Brazil
Toptal member since July 22, 2026
Bio
Caio has spent over 10 years building production-grade AI and data systems across fintech and healthtech. His toolkit centers on Python, LangChain, and AWS Bedrock. While at NVIDIA, Caio rebuilt executive dashboards for capacity planning and cut reporting time by 40%.
Portfolio
.jasc
AI Engineering, LangChain, FastAPI, Azure
Riot Games
Data Engineering, Data Build Tool (dbt), Apache Airflow
Coinbase
Salesforce Apex, CI/CD Automation, LangSmith
Experience
- Python - 10 years
- SQL - 10 years
- Data Engineering - 9 years
- AWS IoT - 7 years
- AI Engineering - 5 years
- Databricks - 5 years
- LangChain - 2 years
- LangGraph - 2 years
Preferred Environment
Azure, Databricks, Apache Airflow, Datadog, Terraform, GitHub Actions
The most amazing...
...AI system I've architected is a HIPAA-compliant multi-agent framework that reduced hallucination rates by 20% for Series A and B startups.
Work Experience
Senior Data & AI Engineer
2024 - PRESENT
.jasc
- Architected a multi-step agentic framework for HIPAA-compliant AI agents on Azure using LangChain and Langfuse, successfully reducing hallucination rates by 20% through precise data-driven adjustments.
- Developed a serverless RAG application with AWS Bedrock and Textract that automated the analysis of complex 200-page documents, saving 15 hours of manual review time weekly for clients in the fintech sector.
- Designed a robust orchestration layer using FastAPI with built-in circuit breakers to safely route tasks among multiple LLM agents, enhancing reliability and performance across SaaS applications.
- Partnered with various Series A/B startups to transform their AI strategies from concept to production-ready solutions while ensuring compliance with industry regulations and standards.
Technologies: AI Engineering, LangChain, FastAPI, Azure
Senior Data Engineer
2026 - 2026
Riot Games
- Revamped executive dashboards by redefining metric semantics and transitioning to cluster-level granularity, enhancing incident attribution accuracy by over 30% for better operational insights.
- Architected a multi-step framework using dbt and Airflow for data transformation processes that improved data pipeline efficiency by 25%, significantly reducing processing time for analytics.
- Integrated DataDog monitoring tools to track performance metrics in real-time, leading to a 15% reduction in downtime through proactive issue identification and resolution.
- Collaborated with cross-functional teams to ensure seamless deployment of data solutions within AWS environments while maintaining compliance with industry standards.
Technologies: Data Engineering, Data Build Tool (dbt), Apache Airflow
Senior AI/ML Engineer
2025 - 2026
Coinbase
- Architected a robust LLM-powered spam classifier for Trust and Safety using Go and Salesforce Apex that enhanced transparency by generating human-readable audit logs for flagged content.
- Implemented advanced LLM Ops primitives utilizing LangSmith to streamline prompt management and chain observability. Integrated with DataDog to enable comprehensive end-to-end production tracing.
- Managed automated CI/CD pipelines through GitHub Actions that ensured resilient model updates and deployment cycles while reducing downtime by 30% during transitions.
Technologies: Salesforce Apex, CI/CD Automation, LangSmith
Senior Data Engineer
2025 - 2026
NVIDIA
- Led a comprehensive overhaul of executive dashboards and datasets that transitioned from provider/region reporting to cluster-level granularity, enhancing capacity attribution by 30% for incident management.
- Designed and maintained hundreds of high-frequency data pipelines on Databricks that processed millions of rows daily. Introduced automated backfill workflows that reduced manual intervention by 40 hours monthly.
- Implemented robust fault-tolerant patterns across all jobs, including timeouts and retries with exponential backoff. This reduced job failure rates by 25%, ensuring higher reliability in data processing.
- Established stringent data quality SLAs using BigEye for schema validation and anomaly detection. Integrated alerts into on-call channels, which improved response times to data issues by 50% within the first month.
- Introduced runbook-first operations with standardized playbooks for failing sources and reprocessing. This streamlined operational efficiency and reduced troubleshooting time by 60% during incidents.
Technologies: Data Engineering, Databricks
Senior AI Consultant
2025 - 2025
Microsoft
- Built a robust agent orchestration service using FastAPI on AKS that supported both streaming and micro-batch endpoints while integrating seamlessly with Azure OpenAI (GPT-4/Phi) via AI Foundry.
- Implemented essential platform primitives for multi-agent routing and tool execution, which included input/output filters and PII scrubbing to ensure compliance with privacy regulations.
- Enhanced system resiliency by incorporating request timeouts, retry mechanisms with backoff strategies, circuit-breakers on upstream model endpoints, and idempotency keys for effective message replays.
- Configured Cosmos DB for managing agent configurations and versions while utilizing Event Hubs for telemetry data collection. Established CI/CD pipelines using GitHub Actions alongside Terraform stacks for efficient deployment processes.
- Developed comprehensive runbooks and quick-start documentation enabling bank teams to self-serve new agents effectively while including safety review checklists to maintain compliance standards.
Technologies: AI Orchestration, CI/CD Pipelines, Terraform
Senior AI Engineer
2024 - 2025
Luma Health
- Launched a dynamic multi-step framework utilizing LangChain that enabled adaptive chatbot responses while integrating Langfuse for real-time monitoring, enhancing error detection capabilities by 40% across the platform.
- Implemented feedback loops with Langfuse’s observability suite that significantly improved GPT-4’s decision-making insights and boosted response accuracy by 20% through targeted data-driven optimizations.
- Streamlined the end-to-end provisioning of LLM resources on Azure using Terraform, which automated model deployments for GPT-4 and Phi-3.5, reducing setup time by an impressive 50% for faster project delivery.
Technologies: AI Engineering, Azure Automation, Real-Time Monitoring, Terraform
Growth Data Engineer
2022 - 2024
Nubank
- Created an AI-driven forecasting tool harnessing Python and statsmodels, now serving as the primary resource for over 1,000 users, including C-level executives, streamlining decision-making processes across the organization.
- Devised a comprehensive data lifecycle protocol that ingested and transformed over 500,000 data points, enhancing the overall quality of insights and enabling the analytics team to respond rapidly to business needs.
- Constructed a scalable data ingestion framework using Clojure and AWS Lambdas, facilitating the ingestion of over 1M data points monthly while optimizing data retrieval processes for analytics teams to improve decision-making capabilities.
Technologies: Python, Artificial Intelligence (AI), Large Language Models (LLMs), SQL, Prompt Engineering, ETL
Fraud Data & AI Engineer
2021 - 2022
ClearSale
- Designed an intuitive NLP TensorFlow application to analyze user sentiment, leading to the identification of key product improvement areas. Insights have driven strategic initiatives that enhanced user engagement by 25% over 6 months.
- Engineered a robust facial recognition system that processed over 10,000 images daily, streamlining the verification process and cutting down fraud detection time by 60%, thus improving customer trust.
- Introduced standard resiliency controls (timeouts, retries, idempotent stages) and monitoring for false‑positive/negative trends.
Technologies: Python, SQL, ETL, Azure, Artificial Intelligence (AI)
Cloud Data & AI Engineer
2020 - 2021
Accenture
- Developed a custom sentiment evaluation process using Python's NLTK, enabling the analysis of over 5,000 customer feedback entries per month. Improved response strategies and customer satisfaction ratings by 25%.
- Constructed an advanced ETL process using Python and Airflow to unify data streams from 10+ sources, enhancing reporting capabilities and empowering analytics teams to derive actionable insights from 1+ million data points.
- Implemented API services on Cloud Run/GKE and standardized CI/CD. Integrated with 3rd‑party/enterprise APIs (including social platforms).
Technologies: Python, Google Cloud Platform (GCP), CI/CD Pipelines, ETL, REST APIs, Artificial Intelligence (AI)
Experience
Multi-agent LLM Orchestration Platform for Enterprise Banking
Designed and shipped V1 of a multi-agent orchestration platform that lets enterprise banking teams stand up their own LLM agents without engineering support.
I built the core service in Python/FastAPI on AKS with streaming and micro-batch endpoints, integrated with Azure OpenAI (GPT-4 and Phi) via AI Foundry. My work covered the platform primitives—multi-agent routing, tool execution, and guardrails including input/output filters, PII scrubbing, and allowlists — plus the resiliency layer: request timeouts, retries with exponential backoff and jitter, circuit breakers on upstream model endpoints, and idempotency keys for message replay.
I wired Cosmos DB for agent configs and versioning and Event Hubs for telemetry, provisioned the full stack with Terraform, and ran CI/CD through GitHub Actions with rolling Kubernetes deploys gated by pytest unit and integration suites. I delivered under a tight timeline, alongside runbooks, safety review checklists, and quick-start docs, so client teams could self-serve new agents.
I built the core service in Python/FastAPI on AKS with streaming and micro-batch endpoints, integrated with Azure OpenAI (GPT-4 and Phi) via AI Foundry. My work covered the platform primitives—multi-agent routing, tool execution, and guardrails including input/output filters, PII scrubbing, and allowlists — plus the resiliency layer: request timeouts, retries with exponential backoff and jitter, circuit breakers on upstream model endpoints, and idempotency keys for message replay.
I wired Cosmos DB for agent configs and versioning and Event Hubs for telemetry, provisioned the full stack with Terraform, and ran CI/CD through GitHub Actions with rolling Kubernetes deploys gated by pytest unit and integration suites. I delivered under a tight timeline, alongside runbooks, safety review checklists, and quick-start docs, so client teams could self-serve new agents.
Education
2017 - 2020
Bachelor's Degree in Information Systems
Universidade De São Paulo - São Paulo, Brazil
Certifications
DECEMBER 2025 - DECEMBER 2028
AWS Certified AI Practitioner
Amazon Web Services
Skills
Libraries/APIs
REST APIs
Tools
Apache Airflow, Terraform, Azure Automation
Languages
Python, SQL
Platforms
AWS IoT, Databricks, LangSmith, Azure, Google Cloud Platform (GCP), Kubernetes, Docker
Frameworks
LangGraph
Paradigms
ETL
Storage
Datadog
Other
AI Engineering, LangChain, FastAPI, CI/CD Automation, AI Orchestration, CI/CD Pipelines, Large Language Models (LLMs), Artificial Intelligence (AI), GitHub Actions, Machine Learning Operations (MLOps), Agent Orchestration, Data Engineering, Data Build Tool (dbt), Salesforce Apex, Real-Time Monitoring, Prompt Engineering, Software Engineering
Collaboration That Works
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
1
Share your needs
Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2
Choose your talent
Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3
Start your risk-free talent trial
Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.
Top talent is in high demand.
Start hiring