
Rafael Avraham
Verified Expert in Engineering
Data Scientist and Developer
São Paulo - State of São Paulo, Brazil
Toptal member since May 22, 2026
Rafael is a lead AI solution architect with more than eight years of experience designing and deploying end-to-end AI, data, and cloud solutions for enterprise clients, including LIV Golf, Brighton, Whirlpool and Colgate. He specializes in LLM integration, RAG systems, AI agents, MLOps, and solution blueprints, with strong expertise in Python, Databricks, AWS, GCP, and Azure. Rafael helps teams integrate fragmented AI capabilities and build scalable foundations for future AI products.
Portfolio
Experience
- Python - 8 years
- Data Science - 8 years
- Machine Learning - 8 years
- AI Architecture - 8 years
- Google Cloud Platform (GCP) - 5 years
- Amazon Web Services (AWS) - 3 years
- AI Agents - 2 years
- AI Engineering - 2 years
Work Experience
Data Scientist - Global Healthcare Products Company
Colgate-Palmolive - GIT
- Developed Python NLP pipelines on GCP Vertex AI to extract product themes, sentiment, and evidence-backed insights from consumer reviews for Colgate-Palmolive.
- Trained and fine-tuned transformer models with PyTorch and Hugging Face for aspect extraction, sentiment classification, and semantic matching, using controlled experiments and separate validation data.
- Built Gemini-powered conversational agents that answered questions about consumer reviews using retrieved evidence, structured responses, and source citations.
- Designed semantic embedding and clustering methods to discover and consolidate product themes, combining sentence transformers, similarity scoring, and trained pair classifiers.
- Benchmarked 188 experiment stages across 46 frozen NLP pipelines, comparing LLM, pretrained, and fine-tuned approaches on extraction, sentiment, grouping, runtime, and memory.
- Analyzed ratings, review trends, and anomalies with Python and SQL, identifying data-quality issues involving duplicated reviews, product mismatches, incomplete snapshots, and SKU linkage.
- Developed market and SKU readiness scoring models using Snowflake and Python, incorporating data coverage, recency, mapping quality, sensitivity analysis, and historical backtesting.
- Created source-linked evaluation datasets and reproducible NLP benchmarks, measuring precision, recall, F1, sentiment accuracy, and theme-grouping quality while checking for data leakage.
- Engineered Snowflake-to-GCP analytics workflows with immutable manifests, checkpoints, caching, and output validation to support traceable AI generation and reliable recovery.
- Modeled AI inference costs and benchmarked local CPU alternatives, analyzing token usage, batch processing, cache reuse, latency, and memory to guide architecture and rollout decisions.
Lead AI Solution Architect
Afya
- Designed AI solution blueprints for production agent systems, including orchestration, tool use, iterative reasoning loops, and multi-agent workflows.
- Enhanced LLMOps observability and evaluation pipelines with automated quality checks, tracing, metrics, and monitoring for safer model behavior.
- Integrated AI capabilities across AWS, GCP, Azure, and Databricks pipelines for training, inference, CI/CD, and production agent execution.
Senior Data Scientist
Uber
- Conducted analysis and created dashboards to inform business decisions.
- Architected and deployed scalable machine learning models to address and resolve key business challenges.
- Designed and analyzed A/B tests to optimize product features and strategies.
- Collaborated with cross-functional teams to translate business needs into data solutions.
- Tracked and maintained model accuracy and effectiveness.
- Documented processes, models, and findings for transparency and reproducibility.
Lead Data Scientist
PMY Group
- Developed end-to-end computer vision pipelines from training to inference.
- Deployed vision models tailored to client needs on GCP, leveraging GCE, Cloud Run, GKE, and Vertex AI Pipelines.
- Trained computer vision models using the latest frameworks, including PyTorch, TensorFlow, and YOLO.
MLOps Tech Lead
Whirlpool
- Developed serverless machine learning architecture on GCP that reduced project infrastructure costs by 99.65% while keeping the system scalable and maintainable.
- Built and deployed machine learning pipelines using Vertex AI, GKE, BigQuery, Cloud Storage, Cloud Functions, Cloud Build, Kedro, and Kubeflow Pipelines.
- Delivered retail search and predictive modeling solutions using Python, SQL, and cloud-native MLOps practices across GCP-based environments.
Data Scientist
Thinkseg
- Led data science and AI architecture work at Thinkseg, growing from intern to head of data science while using Python, SQL, and ML tools to turn raw data into business insights.
- Designed AI/ML solutions for churn prediction, customer clustering, active-client prediction, and segmentation based on customer information and driving behavior.
- Delivered dashboards, BI reports, and KPI tracking for managers and teams, while mentoring data science interns and shaping early AI architecture standards.
Project History
Telegram RAG Agent
https://github.com/rafaelgildin/telegram-rag-agentPenalty Trainer Computer Vision System
https://github.com/rafaelgildin/penalty_projectYOLOv8 People Tracking and Counting
https://github.com/rafaelgildin/track_and_count_peopleEducation
Bachelor's Degree in Computer Science
Mackenzie Presbyterian University - São Paulo, Brazil
Certifications
Google Cloud Certified – Professional Machine Learning Engineer
Google Cloud
Google Cloud Certified Professional Data Engineer
Google Cloud
Google Cloud Certified Professional Machine Learning Engineer
Google Cloud
Google Cloud Certified Associate Cloud Engineer
Google Cloud
Skills
Libraries/APIs
OpenAI API
Languages
Python
Platforms
Databricks, Amazon Web Services (AWS), Google Cloud Platform (GCP), Kubernetes, Docker
Industry Expertise
Marketing
Other
AI Architecture, RAG Systems, AI Agents, AI Engineering, Data Science, Machine Learning, Large Language Model Operations (LLMOps), Artificial Intelligence (AI), LangChain, PostgreSQL 9, Data Engineering, Deep Learning
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring