Carlos is currently unavailable
Carlos Dutra
Verified Expert in Engineering
Software Developer
São Paulo - State of São Paulo, Brazil
Toptal member since October 6, 2020
Bio
Carlos is an AI and ML expert with 10+ years of experience, recently focused on data science, generative AI, agent development, and computer vision. He has built scalable LLM-based systems and led multicultural teams across different countries. His work spans risk, fraud, marketing, and education, blending technical depth with strategic leadership.
Portfolio
Thats Great News, LLC
Python, Artificial Intelligence (AI), System Architecture, Source Code Review...
Typeform SL
Python, Machine Learning, Data Science, Amazon Bedrock...
Everyday AI, Inc.
LiveKit, Python, Node.js, Artificial Intelligence (AI)
Experience
- Amazon Web Services (AWS) - 10 years
- Python - 10 years
- Machine Learning - 10 years
- Microsoft Excel - 10 years
- Data Science - 8 years
- Retrieval-augmented Generation (RAG) - 3 years
- Large Language Models (LLMs) - 3 years
- PySpark - 2 years
Preferred Environment
Visual Studio, Linux, Artificial Intelligence (AI), AWS Command Line Interface (CLI), LangGraph
The most amazing...
...project I've implemented was a system of recommendations and offers for an application based on multi-armed bandit, increasing the sales of independent sellers.
Work Experience
AI Systems Audit and Architecture Review Engineer
2026 - PRESENT
Thats Great News, LLC
- Audited a legacy AI lead-generation system built by an external agency, surfacing recurring crashes, opaque costs, pipeline inefficiencies, and multiple critical security vulnerabilities.
- Helped define the forward-looking AI strategy, aligning model choices, pipeline redesign, and cost controls with the team's scale targets and day-to-day operational realities.
- Supported the migration of cloud infrastructure and SaaS tooling from the former agency's accounts into environments fully owned and controlled by the client's own organization.
Technologies: Python, Artificial Intelligence (AI), System Architecture, Source Code Review, Document Parsing, Optical Character Recognition (OCR), PDF, Image Processing, Large Language Models (LLMs), API Integration, Amazon Web Services (AWS), Model Evaluation, Cost Reduction & Optimization (Cost-down), Natural Language Processing (NLP), Amazon S3 (AWS S3), AWS Lambda, Cloud Architecture, RAG Pipelines, Vector Databases
AI/ML Engineer
2026 - PRESENT
Typeform SL
- Developed a multi-layer evaluation framework combining deterministic validation, LLM-as-judge assessment, and API dry-run testing across test cases, with MLflow integration for experiment tracking and priority-based pass rate thresholds.
- Defined multi-agent system architecture using Amazon Bedrock AgentCore and A2A protocol, and MCP, implementing supervisor-based orchestration with LangGraph for LLM routing and streaming inter-agent communication across specialized domain agents.
- Implemented dual observability infrastructure with DataDog for real-time APM and OpenTelemetry for long-term trace storage in S3, capturing LLM interactions, token usage, and graph execution across all AI agents.
Technologies: Python, Machine Learning, Data Science, Amazon Bedrock, Amazon Web Services (AWS), Agentic AI, Artificial Intelligence (AI), LangGraph, Arize
AI Voice Engineer
2026 - 2026
Everyday AI, Inc.
- Developed a full-stack voice assistant using Next.js 15, React 19, Python, and LiveKit WebRTC, integrating GPT-4o-mini, Cartesia TTS, and AssemblyAI STT to deliver real-time conversational experiences for elderly care applications.
- Architected a modular skill system with PostgreSQL/Prisma back end, enabling medication reminders, trivia games, and news delivery, reducing feature development time through reusable command patterns and type-safe APIs.
- Implemented back-end-authoritative state management for interactive modules, eliminating client-server race conditions and improving session reliability across unreliable network conditions.
Technologies: LiveKit, Python, Node.js, Artificial Intelligence (AI)
Generative AI Developer
2025 - 2026
LotLinx, Inc
- Developed an AI-powered sales assistant for car dealerships using retrieval-augmented generation (RAG), integrating sales, market, and vehicle attribute data from APIs, internal databases, and web scrapers, with search powered by vector databases.
- Built a GenAI-based vehicle image enhancement system, using vision models and computer vision techniques.
- Fine-tuned and deployed customized models for image processing, using cutting-edge techniques.
Technologies: Artificial Intelligence (AI), Generative Design, Generative Pre-trained Transformers (GPT), Image Processing, AIOps, Image Generation, Computer Vision, Retrieval-augmented Generation (RAG), Agentic AI, Architecture, GPU Computing, Large Language Model Operations (LLMOps)
Data Scientist
2025 - 2025
DFS Dashboard LLC - Main
- Implemented 23+ optimization algorithms using linear programming, Monte Carlo simulation, and probabilistic modeling to generate DraftKings-compliant lineups.
- Designed and deployed FastAPI microservice with PostgreSQL connection pooling and multi-environment AWS ECS deployment.
- Created GitHub Actions CI/CD pipeline for automated ECR/ECS deployments with health checks.
Technologies: Data Science, Statistics, Fantasy Sports
AI Agents Developer
2025 - 2025
Javier Morales
- Designed the architecture and phased migration plan to transition from a no-code Uchat platform to a fully customizable, self-hosted multi-agent system using LangGraph, enabling deeper customization and greater scalability.
- Redesigned the logic of existing AI agents used for lead qualification and scheduling, identifying and fixing major issues in the current implementation.
- Evaluated multiple no-code and low-code platforms to rapidly prototype AI agent features, accelerating product validation and user feedback collection with minimal engineering effort in the early development phase.
Technologies: AI Agents, Python, Artificial Intelligence (AI), OpenAI, n8n, GoHighLevel, Uchat, Agentic AI, Architecture
ChatGPT AI Content Improvement Specialist
2025 - 2025
Arbel Growth partners - Main
- Developed AI agents using deep learning and LLMs (OpenAI) to produce high-quality e-learning documents across various domains, ensuring consistency, reducing hallucinations, and maintaining top technical quality.
- Expanded the solution to generate full lesson videos—including slides, narration, text-to-speech, and quality assessment—to transform written content into engaging video formats.
- Deployed heavy pipelines with rigorous logging, debugging, parallelization, custom models, and optimal GPU usage, ensuring scalable and reliable system operations.
Technologies: ChatGPT, Generative Pre-trained Transformers (GPT), OpenAI GPT-4 API, Machine Learning, Artificial Intelligence (AI), LangChain, Large Language Models (LLMs), Computer Vision, Retrieval-augmented Generation (RAG), Text-to-Speech (TTS), PyTorch, Architecture, GPU Computing, Large Language Model Operations (LLMOps)
Data Science Manager
2024 - 2025
Trustly
- Led income prediction, balance modeling, and bank activity categorization using deep learning, LLMs, BERT, and finetuning to enhance financial insights.
- Hired four data scientists and expanded the team to drive deep learning projects for advanced data-driven analysis.
- Delivered the company’s first LLM-powered chatbot, utilizing deep learning to improve customer engagement and digital innovation.
Technologies: Python, LangGraph, LangChain, AWS Command Line Interface (CLI), Multimodal GenAI, Chatbots, Risk Models, OpenAI, Amazon Bedrock, Large Language Models (LLMs), Retrieval-augmented Generation (RAG), Fantasy Sports, Agentic AI, PyTorch, Architecture
GPT Wrapper Application Developer
2024 - 2024
Hapily Inc.
- Designed a streamlined UI that coordinates multiple LangChain agents for prompt management and data retrieval, demonstrating how AI agents can collaboratively assist users in marketing data tasks through a smooth, interactive interface.
- Built a retrieval-augmented generation (RAG) workflow leveraging LangChain agents to seamlessly integrate GPT with HubSpot’s API, enabling real-time data querying and dynamic CRM updates in the PoC.
- Implemented advanced prompt engineering within the GPT wrapper to intelligently guide user input and recommend relevant objects, fields, and sample data, enhancing the AI-driven user experience.
Technologies: Generative Pre-trained Transformers (GPT), Artificial Intelligence (AI), OpenAI GPT-4 API, Python, AI Agents, Retrieval-augmented Generation (RAG)
Senior Staff Data Scientist | Senior Manager
2020 - 2024
Wildlife Studios
- Developed an AI assistant based on LLMs to help business analysts perform analysis in Looker, the enterprise BI tool.
- Created and managed the company's first business intelligence team for marketing and user acquisition.
- Handled the data for the company's biggest game, Sniper 3D, improving experimentation, analysis, and statistical modeling dynamics, supporting the team in achieving a 20% increase in revenue.
Technologies: Python, Large Language Models (LLMs), Data Science, Amazon Web Services (AWS), Retrieval-augmented Generation (RAG), Machine Learning Operations (MLOps), Predictive Modeling, PySpark, Statistics, C#, Data Analysis, Data Mining, AI Chatbots, Databases, Generative Artificial Intelligence (GenAI), ChatGPT, OpenAI, Natural Language Processing (NLP), Data Warehousing, Unstructured Data Analysis, Deep Learning, Convolutional Neural Networks (CNNs), Chatbots, Image Processing, Data Engineering, Digital Signal Processing, Agentic AI, PyTorch, Architecture
CTO | Co-founder
2017 - 2021
MeuVendoo
- Deployed a highly scalable intelligent alert engine, processing over 1 million transactions registered in the application.
- Developed a crawler using Selenium to capture and structure price information from hundreds of product catalogs.
- Integrated with several financial and retail partners through RESTFul APIs.
Technologies: Amazon Web Services (AWS), Microsoft Power BI, Firebase, ETL, DevOps, GitLab CI/CD, Agile, MySQL, AWS Lambda, Apache Airflow, Python, Android, Java, BigQuery, Data Analysis, Data Mining, Data Scraping, Databases, REST APIs, CTO, Computer Vision, Convolutional Neural Networks (CNNs), Data Engineering, API Integration, Architecture
Senior Data Scientist
2017 - 2018
Porto Seguro
- Developed a dynamic pricing system using machine learning, applying multi-armed bandit modeling and providing a service in AWS Lambda.
- Built a system that recommended products in an automatic telephone service (IVR) based on XGBoost and using cadastral, behavioral, and credit card features.
- Created dashboards and automatic reports in Google Data Studio for operation telemetry, following the main indicators.
Technologies: Amazon Web Services (AWS), XGBoost, Microsoft Excel, Oracle Database, MySQL, R, Python, Data Science, Machine Learning Operations (MLOps), Predictive Modeling, Data Analysis, Data Mining, Databases, Data Warehousing
Risk Analyst
2014 - 2015
Banco Bradesco
- Developed studies to manage the bank's credit policies.
- Monitored the indicators of default and profitability of the credit portfolio.
- Oversaw performance of statistical models and defined policies and product cutoff points for companies and individuals.
Technologies: Clustering, Decision Trees, Logistic Regression, Microsoft Excel, SQL, SAS, Data Science, Python, Predictive Modeling, Data Analysis, Data Mining, Databases, Data Warehousing
Research and Development Analyst
2009 - 2014
Kron Electrical Instruments
- Developed an automation system for the testing of products, increasing productivity by about five times, and achieving higher reliability and accuracy.
- Wrote MATLAB, C, VBA, and Python code to perform electrical engineering simulations such as a fast Fourier transform and interpolations.
- Obtained approval in external laboratory tests, according to the standards of electrical measurement in universities, research centers, and regulatory agencies.
Technologies: Java, Assembly, C, Scilab, Octave, MATLAB, Python, Data Analysis, Data Mining, Databases, Digital Signal Processing
Experience
LotGPT, AI Advisor Tailored for Car Dealerships
https://lotlinx.com/lotgpt/Senior member of technical staff contributing to LotGPT, the first dealer-facing conversational AI at LotLinx. Worked on key components of the LLM architecture that turn VIN-level, shopper-behavior, and market data into real-time pricing, merchandising, and aging-risk insights. Led the development of the AI vehicle image enhancer as an agent tool and contributed to AI avatar generation features for dealer workflows.
Lineup Optimizer for Fantasy Sports Gaming
https://dfsdashboard.com/I built a production FastAPI back end for DFS Dashboard - an NFL DraftKings
lineup optimization platform serving daily fantasy sports players.
Technical implementation:
• Designed and deployed FastAPI microservice with PostgreSQL connection pooling and multi-environment AWS ECS deployment.
• Implemented 23+ optimization algorithms using linear programming, Monte Carlo simulation, and probabilistic modeling to generate DraftKings-compliant lineups.
• Integrated the Odds API, weather data, and Vegas lines for real-time player projections.
• Created GitHub Actions CI/CD pipeline for automated ECR/ECS deployments with health checks.
• Optimized database queries with connection pooling, deduplication, and slate-specific filtering.
• Established code quality standards with pre-commit hooks and structured logging.
The system manages complex multi-constraint optimization across salary caps, positions, stacking rules, and exposure limits.
lineup optimization platform serving daily fantasy sports players.
Technical implementation:
• Designed and deployed FastAPI microservice with PostgreSQL connection pooling and multi-environment AWS ECS deployment.
• Implemented 23+ optimization algorithms using linear programming, Monte Carlo simulation, and probabilistic modeling to generate DraftKings-compliant lineups.
• Integrated the Odds API, weather data, and Vegas lines for real-time player projections.
• Created GitHub Actions CI/CD pipeline for automated ECR/ECS deployments with health checks.
• Optimized database queries with connection pooling, deduplication, and slate-specific filtering.
• Established code quality standards with pre-commit hooks and structured logging.
The system manages complex multi-constraint optimization across salary caps, positions, stacking rules, and exposure limits.
AI Assistant for Hubspot (CRM) Management
Developed an AI demo assistant using LLMs and integrated it with Hubspot API to assist in creating new accounts through an interface 100% based on natural language. I developed the project from end to end, including the AI , DevOps, IaaS, front end, back end, and integration components using Python, LangChain, OpenAI API, AWS, and Streamlit using a serverless architecture.
MeuVendoo App
I developed an application to manage the work of individual sellers in a door-to-door model. It has elements of financial management (CRM and ERP) and an intelligence module that notifies the user of the best moment and the required action, such as offering products to the end customer.
Structuring the Brazilian Social and Demographic Data for Credit Analysis
I developed a Python module for capturing, structuring, and building a feature store based on public data provided by the Brazilian Institute of Geography and Statistics (IBGE) of the various national household sample surveys (PNAD), with information on income and unemployment of the population, among others.
Education
2017 - 2021
Master's Degree in Applied Mathematics
Universidade de São Paulo - São Paulo, SP, Brazil
2012 - 2016
Bachelor's Degree in Applied and Computational Mathematis
Universidade de São Paulo - São Paulo, SP, Brazil
Certifications
MAY 2026 - PRESENT
Leading the AI-Driven Organization
MIT Sloan School of Management
JANUARY 2026 - PRESENT
Hugging Face Diffusers Contributor - MVP Program
Hugging Face
MAY 2025 - PRESENT
Fundamentals of MCP
Hugging Face
OCTOBER 2024 - PRESENT
Introduction to LangGraph
LangChain Academy
AUGUST 2021 - PRESENT
Architecting on AWS
Amazon Web Services
AUGUST 2021 - PRESENT
Deep Learning on AWS
Amazon Web Services
Skills
Libraries/APIs
PySpark, REST APIs, XGBoost, PyTorch, Node.js
Tools
Microsoft Excel, Visual Studio, Apache Airflow, ChatGPT, GitLab CI/CD, MATLAB, Scilab, Microsoft Power BI, BigQuery, AWS Command Line Interface (CLI), n8n, AWS ELB
Languages
Python, R, SQL, C#, Java, SAS, Octave, C, Assembly
Platforms
Linux, LiveKit, Oracle Database, Amazon Web Services (AWS), Android, AWS Lambda, Firebase, AWS Cloud Computing Services, GoHighLevel
Storage
Databases, MySQL, Amazon S3 (AWS S3)
Frameworks
LangGraph
Paradigms
Agile, DevOps, ETL, Model Context Protocol (MCP)
Other
Linear Algebra, Optimization, Statistics, Machine Learning, Digital Signal Processing, Data Science, Artificial Intelligence (AI), Retrieval-augmented Generation (RAG), Machine Learning Operations (MLOps), Predictive Modeling, Data Analysis, Data Mining, Natural Language Processing (NLP), Data Warehousing, Unstructured Data Analysis, CTO, Deep Learning, Computer Vision, Chatbots, Data Engineering, API Integration, Agentic AI, Text-to-Speech (TTS), Architecture, Executive Consulting, AI-generated Video, Econometrics, Large Language Models (LLMs), Data Scraping, AI Chatbots, Generative Artificial Intelligence (GenAI), OpenAI, Convolutional Neural Networks (CNNs), Image Processing, Fantasy Sports, GPU Computing, Large Language Model Operations (LLMOps), Time Series Analysis, Logistic Regression, Decision Trees, Clustering, SVMs, Neural Networks, Regression Modeling, Generative Pre-trained Transformers (GPT), OpenAI GPT-4 API, Serverless, Technology, LangChain, Multimodal GenAI, Risk Models, Amazon Bedrock, Generative Design, AIOps, Image Generation, AI Agents, Uchat, Statistical Methods, FastAPI, Diffusion Models, Videos, Arize, System Architecture, Source Code Review, Document Parsing, Optical Character Recognition (OCR), PDF, Model Evaluation, Cost Reduction & Optimization (Cost-down), Cloud Architecture, RAG Pipelines, Vector Databases, Organizational Strategy
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