Luca is currently unavailable
Luca Cerone
Verified Expert in Engineering
ML Engineer | Data Scientist and Developer
Barcelona, Spain
Toptal member since August 28, 2026
Bio
Luca is a data scientist and ML engineer with 11 years of experience delivering end-to-end data products for companies like Bumble, letgo (OLX Group), Domestika, and King (Microsoft). He prioritizes good engineering practices to deploy reliable models across AWS, GCP, and on-premise environments. By combining his robust ML deployment background with active training in generative AI, including RAG and LLMs, Luca helps companies safely implement modern AI to solve concrete business needs.
Portfolio
Domestika
Python, Docker, MySQL, Apache Airflow, Amazon Web Services (AWS), SQL...
Bumble
Python, PyTorch, Scikit-learn, XGBoost, LightGBM, FastAPI, Flask, Kubernetes...
letgo (OLX Group)
Python, Amazon Web Services (AWS), Docker, CI/CD Pipelines, Machine Learning...
Experience
- Programming - 8 years
- A/B Testing - 8 years
- Python - 8 years
- Machine Learning - 8 years
- Kubernetes - 5 years
- Machine Learning Operations (MLOps) - 5 years
- Deep Learning - 5 years
- Kubeflow - 2 years
Work Experience
Senior Data Scientist
2024 - 2025
Domestika
- Built and deployed 21 in-house similar courses recommendation models using item2vec embeddings, effectively replacing a costly 3rd-party model.
- Conducted an A/B test for the new course recommendation models that demonstrated a 78% increase in the click-through rate (CTR) and a 12% increase in course purchases.
- Engineered and maintained Apache Airflow pipelines on AWS to orchestrate the periodic, automated retraining of the 21 regional models and integrate recommendations into a MySQL database.
- Packaged machine learning models as isolated Docker images to guarantee consistent and reliable deployments across AWS cloud environments.
- Optimized the user recommendation model using signals like course visits and purchases, improving recommendation CTR by 12% and driving a 7% increase in course purchases from the homepage.
- Designed search tracking infrastructure and developed reporting platforms in Amplitude and Looker to deeply analyze user search behavior across the catalog.
- Analyzed user behavior to estimate customer lifetime value (LTV) and identified a core North Star metric for Domestika's subscription model.
- Spearheaded the broader introduction of machine learning models into the Domestika product, consistently aligning data science projects with company OKRs and main KPIs.
Technologies: Python, Docker, MySQL, Apache Airflow, Amazon Web Services (AWS), SQL, Recommendation Systems, Amplitude, Amazon S3 (AWS S3), AWS Glue, Amazon Athena, Amazon Redshift, Amazon Redshift Spectrum, Amazon EKS, Amazon Elastic Container Service (ECS), Artificial Intelligence (AI), Large Language Models (LLMs), Software Engineering, Data Analysis, Solution Architecture
Senior Machine Learning Scientist
2022 - 2024
Bumble
- Refactored the codebase of a massive-scale graph neural network match prediction model, achieving a 10x memory reduction and a 75% decrease in training times.
- Implemented Kubeflow pipelines to support the global rollout of the match prediction model, enabling daily embeddings refresh, parallel model training, and reduced prediction latency.
- Powered a core paid feature using the optimized match prediction model, driving a +4% increase in revenue alongside measurable improvements in user retention and match metrics.
- Developed an image content extractor leveraging zero-shot CLIP capabilities and engineered prompts, which won an internal data science hackathon.
- Analyzed user profiles using the custom image extraction model to directly inform product design decisions, such as improving the interests taxonomy and identifying profiles without visible faces.
- Engineered and deployed an end-to-end automated pipeline to daily train a fraud detection model, alongside a secure API to serve predictions in production.
- Designed enhancements to the recommendation platform's architecture by proposing vector embedding databases like Milvus to optimize the encounters queue for competing business metrics.
- Supported the development of continuous delivery and service reliability within an on-premises Kubernetes environment by adopting a GitOps culture and utilizing Argo CD.
- Guided internal training sessions on clean code principles for data scientists, significantly improving the maintenance, readability, and testability of machine learning projects.
Technologies: Python, PyTorch, Scikit-learn, XGBoost, LightGBM, FastAPI, Flask, Kubernetes, Kubeflow, Argo CD, KServe, GitOps, Docker, CI/CD Pipelines, Google Cloud, Milvus, SQL, Graph Neural Networks (GNNs), Multimodal Models, Recommendation Systems, Fraud Detection, Clean Code, GitHub Actions, Artificial Intelligence (AI), Software Engineering, Data Analysis, Solution Architecture
Senior Data Scientist
2017 - 2022
letgo (OLX Group)
- Developed a blender in the main feed to optimize impressions based on user preferences and business rules, driving an increase in visits to car listings and an increase in calls to the car sales team.
- Architected automated pipelines on AWS to train, evaluate, and deploy a multimodal listing classifier (text and images) for the US and Turkish markets, which improved accuracy and increased buyer-seller contacts.
- Drove user personalization initiatives by customizing verticals based on inferred user preferences, achieving consistent uplifts in conversion rates across multiple iterations.
- Deployed an item-to-item recommender system specifically for retention campaigns that successfully drove a 10% increase in open rates.
- Created car image quality and vehicle orientation detection computer vision models to increase catalog liquidity and lead to professional sellers, earning a feature on letgo's technical blog.
- Designed and implemented CI/CD pipelines using Jenkins, Docker, and pytest for internal data science tools, ensuring rapid and safe code delivery.
- Initiated an automated monitoring system to ensure data quality and trigger timely alerts in response to data drift or pipeline degradation.
- Mentored junior data scientists, guiding them on scientific and technical best practices and successfully integrating them into letgo's Agile processes.
Technologies: Python, Amazon Web Services (AWS), Docker, CI/CD Pipelines, Machine Learning, Machine Learning Operations (MLOps), A/B Testing, PyTorch, TensorFlow, Flask, Unit Testing, Natural Language Processing (NLP), Computer Vision, Deep Learning, Recommendation Systems, Amazon SageMaker, Amazon S3 (AWS S3), Jenkins, Amplitude, Tableau, ECS, Git, EC2, Artificial Intelligence (AI), Software Engineering, Data Analysis, Solution Architecture
Data Scientist
2014 - 2017
King (Activision/Microsoft)
- Analyzed large-scale player behavior data to optimize in-game mechanics, directly supporting game and level designers in maximizing user engagement and appeal.
- Designed and executed rigorous A/B tests on new game features, reporting statistical results to studio leadership to guide monetization and retention decisions.
- Spearheaded a multi-month series of rigorous A/B tests and statistical analyses that resolved strategic debates, successfully validating the business value of LiveOps and driving its full adoption across all King game titles.
- Engineered automated ETL pipelines to feed custom-built dashboards, providing reliable, continuous operational metrics to cross-functional stakeholders.
- Developed custom analytical tools and internal R packages that streamlined data analysis workflows across the organization.
- Earned two internal data science guild awards for building open-source-style internal packages that significantly benefited the wider data science community.
Technologies: R, Python, SQL, Big Data, Spark, A/B Testing, ETL, Statistical Analysis, Data Visualization, EC2, Data Analysis
Project History
Vehicle Orientation Model
I built a computer vision model for letgo to automatically detect vehicle orientation and score car image quality, aiming to increase catalog liquidity and drive more leads to professional sellers. The project was highly successful, leading to a featured article on the company's tech blog outlining the approach: if you want to sell your car, you should take a side picture first.
Education
2009 - 2013
PhD in Bioinformatics and Systems Biology
University College Dublin (UCD) - Dublin, Ireland
2005 - 2007
Master's Degree in Mathematics
Sapienza University of Rome - Rome, Italy
2001 - 2005
Bachelor's Degree in Mathematics
Sapienza University of Rome - Rome, Italy
Certifications
JUNE 2026 - PRESENT
MCP: Build Rich-context AI Apps with Anthropic
Deeplearning.ai
JUNE 2026 - PRESENT
Retrieval Augmented Generation (RAG)
Deeplearning.ai
JUNE 2026 - PRESENT
Agentic AI
Deeplearning.ai
MAY 2023 - PRESENT
Reinforcement Learning Specialization
Coursera
MAY 2022 - PRESENT
Introduction to Containers w/ Docker, Kubernetes & OpenShift
IBM
MARCH 2022 - PRESENT
Machine Learning Data Lifecycle in Production
Coursera
FEBRUARY 2022 - PRESENT
Machine Learning in Production
Deeplearning.ai
FEBRUARY 2020 - MARCH 2023
AWS Certified Solutions Architect - Associate
Amazon Web Services (AWS)
JANUARY 2018 - PRESENT
Deep Learning Nanodegree Foundation
Udacity
Skills
Libraries/APIs
Scikit-learn, XGBoost, PyTorch, TensorFlow
Tools
Apache Airflow, Jenkins, TeamCity, Amazon SageMaker, AWS Glue, Amazon Athena, Amazon Redshift Spectrum, Amazon EKS, Amazon Elastic Container Service (ECS), Tableau, Git
Languages
Python, SQL, R
Paradigms
ETL, Unit Testing, Clean Code, Model Context Protocol (MCP)
Frameworks
LightGBM, Flask, Spark
Platforms
Kubernetes, Kubeflow, Amazon Web Services (AWS), Docker, KServe, AWS Lambda
Storage
MySQL, Amazon S3 (AWS S3), Google Cloud
Other
Machine Learning Operations (MLOps), Deep Learning, Statistics, Machine Learning, Programming, A/B Testing, Statistical Analysis, Natural Language Processing (NLP), Recommendation Systems, Artificial Intelligence (AI), Data Analysis, Solution Architecture, GitOps, CI/CD Pipelines, Milvus, Big Data, Data Visualization, FastAPI, Multimodal Models, Fraud Detection, Amplitude, Agentic AI, Large Language Models (LLMs), Software Engineering, Generative Artificial Intelligence (GenAI), GCP, GitHub Actions, EC2, ECS, EMR, Computer Vision, Argo CD, Graph Neural Networks (GNNs), Reinforcement Learning, Deep Reinforcement Learning, Retrieval-augmented Generation (RAG), AI Agents, Amazon Redshift
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