Francis Marti
Verified Expert in Engineering
Machine Learning Engineer and Developer
Saint Petersburg, FL, United States
Toptal member since June 30, 2025
Francis is an experienced software engineer and machine learning specialist with 20+ years of experience in the tech industry. He has delivered machine learning energy, finance, and insurance solutions, leveraging Python, TensorFlow, XGBoost, FastAPI, and distributed systems. Francis is an expert in large language models (LLMs) integration and real-time inference at scale.
Portfolio
Experience
- Linux - 20 years
- Python - 7 years
- TensorFlow - 7 years
- Pandas - 7 years
- Deep Learning - 7 years
- Pytest - 7 years
- FastAPI - 6 years
- LangChain - 2 years
Preferred Environment
Python, Pandas, TensorFlow, Jupyter, FastAPI, PostgresDB, CockroachDB, LangChain, OpenAI, Ollama, Large Language Models (LLMs), Machine Learning
The most amazing...
...training pipeline and model rewrite I've implemented, after finding a data transformation bottleneck, cut runtime significantly, saving over $1 million.
Work Experience
Machine Learning Engineer
Self-employed
- Built a multi-tenant app using GenAI and LLMs to extract structured data from unstructured email attachments (PDF, HTML, text), enabling automated processing and integration into downstream business workflows.
- Enabled an interface for specialists to define, review, and approve new documents, resulting in a five-fold productivity boost through streamlined validation and automation.
- Captured data at each processing stage to track and aggregate invoice amounts, enabling audit log reviews for transparency and compliance.
Machine Learning Engineer
Self-employed
- Engineered and tested high-volume OLTP/OLAP queries on distributed systems, including CockroachDB and SingleStore.
- Built two terabytes of synthetic datasets and delivered insights using Python, Jupyter, Pandas, and Seaborn.
- Implemented change data capture (CDC) from MongoDB, using Apache Kafka for real-time data streaming.
Machine Learning Engineer
AIM Consulting
- Worked for TC Energy, generating operational metrics to forecast oil and gas supply and demand across North American pipelines.
- Refactored Jupyter workflows into a modular, production-grade Python codebase for training and inference reuse.
- Designed a config-driven ML pipeline using YAML, Pandas, Scikit-Learn, XGBoost, and Parquet.
- Cut Amazon EC2 usage from six to one instance, boosting compute efficiency 48 times and inference speed tenfold.
- Automated ML model deployment with Docker, ECS, and AWS CloudFormation, reducing costs by 144% and enabling CI/CD.
- Drove over $1 million in annual savings through system performance tuning and end-to-end ML automation.
Data Scientist and Machine Learning Engineer
J.P. Morgan Corporate & Investment Bank
- Proposed modernization of rules-based processing for bank-to-bank SWIFT messages to improve automation and accuracy.
- Spearheaded the development of a CNN-based vision model using Python, TensorFlow, and Keras, boosting accuracy from 81% to 97%.
- Evaluated multiple models and selected the CNN model, providing classification with confidence scores for straight-through processing.
- Led a team of two colleagues to harden and productionize the app with REST APIs using Python, FastAPI, Docker, Kubernetes, and Red Hat Linux.
Experience
Oil and Gas Predictive Modeling Project for TC Energy
With leadership approval, I led a full rewrite of the system, streamlining data extraction, transformation, model training, evaluation, and deployment. The prior manual process—training separate models for each pipeline station—was replaced with an end-to-end automated workflow. I optimized infrastructure, reducing Amazon EC2 usage from six instances (48 cores) to just one, and eliminated the need for AWS Lambda and Step Functions. Deployment time dropped from three weeks with three data scientists to just 2.5 hours. These improvements resulted in significant gains in speed, maintainability, and cost-efficiency, with estimated annual AWS savings of $1 million.
Skills
Libraries/APIs
Pandas, TensorFlow, Scikit-learn, XGBoost, REST APIs
Tools
Jupyter, Plotly, Pytest, ChatGPT, Seaborn, Amazon SageMaker, Kafka Connect
Languages
Python, Java, SQL, C, GraphQL
Frameworks
JUnit 5, Spark, Flask, AWS HA
Paradigms
Automation, Object-oriented Programming (OOP), ETL, Event-driven Architecture, DevOps, Event-driven Programming
Platforms
Ollama, Linux, Docker, Oracle, Apache Kafka, Kubernetes, AWS Lambda, Amazon Web Services (AWS), Azure
Storage
Data Pipelines, MySQL, CockroachDB, MongoDB, Amazon DynamoDB, PostgreSQL, NoSQL, Google Cloud
Other
FastAPI, LangChain, OpenAI, RESTFul APIs, Deep Learning, Convolutional Neural Networks (CNNs), Large Language Models (LLMs), Generative Artificial Intelligence (GenAI), Vector Databases, Retrieval-augmented Generation (RAG), APIs, Full-stack, Agentic AI, Back-end, Artificial Intelligence (AI), Software Architecture, AI Agents, Dashboards, Data Engineering, Data Modeling, PostgresDB, SingleStore, Parquet, Auroa, Pub/Sub, Azure Compute Services, Natural Language Processing (NLP), Machine 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