Pedro is available for hire
Hire PedroPedro Lima
Verified Expert in Engineering
Software Developer
Porto, Portugal
Toptal member since September 27, 2017
Bio
Pedro is a software developer and architect specializing in data science, machine learning, and AI. He has extensive experience in the end-to-end process of conceiving, designing, developing, and deploying data applications for large companies and startups.
Portfolio
The ROIG Group LLC
Document Processing, Finance, Data Pipelines, AI Agents, AIOps
Inselspital (Bern University Hospital)
Large Language Models (LLMs), Artificial Intelligence (AI), Python...
Yakoa Consulting Limited
OpenAI, Python, Artificial Intelligence (AI), Large Language Models (LLMs)...
Experience
- Machine Learning - 10 years
- Python - 10 years
- Data Science - 10 years
- Text Mining - 7 years
- Natural Language Processing (NLP) - 7 years
- Data Pipelines - 6 years
- Chatbots - 5 years
- Agentic AI - 3 years
Preferred Environment
Jupyter, Python, MacOS, Linux, Git, Docker, LangChain, PyTorch, Codex, Claude Code
The most amazing...
...thing I've developed was an AI agent that learned to play text-based games and won the Microsoft TextWorld competition.
Work Experience
AI Engineer - Intelligent Document Processing for Finance
2025 - PRESENT
The ROIG Group LLC
- Co-designed the agentic architecture behind the document intelligence platform, defining agent roles, tool boundaries, retrieval strategy and evaluation criteria.
- Developed the knowledge layer end to end: ingestion, consolidation, retrieval, tuning and integrations with data sources.
- AI enablement. Helped grow the team from two builders to all builders. Mentored business experts on building with AI, shortening the loop from "requirements discussed" to "working prototype delivered".
Technologies: Document Processing, Finance, Data Pipelines, AI Agents, AIOps
AI for Neurosurgery
2026 - 2026
Inselspital (Bern University Hospital)
- Designed a novel generative approach to building a clinical taxonomy from scratch.
- Built the evaluation and refinement loop that made the taxonomy trustworthy. An automated quality harness that scored every generation run, isolated failure modes, and fed fixes back into the generation strategy.
- Oversaw the application of the tool across the full taxonomy programme.
Technologies: Large Language Models (LLMs), Artificial Intelligence (AI), Python, Machine Learning, AI Pipeline, AI Agents
Senior LLM/RAG Engineer (Legal documents)
2025 - 2026
Yakoa Consulting Limited
- Designed and implemented the ingestion pipeline turning legal documents across PDF/scanned PDF/Word/images, into clean, structure-preserving Markdown, plus an automated translation stage producing English versions of documents in other languages.
- Built the retrieval system end to end using a Pinecone vector database, and a search interface wired into ChatGPT.
- Exported the document corpus as a curated, agent-ready file store, published to a shared folder for use with agentic tools (Claude Code, Codex) providing two access patterns: semantic search and file navigation and grep.
Technologies: OpenAI, Python, Artificial Intelligence (AI), Large Language Models (LLMs), Retrieval-augmented Generation (RAG), Pinecone, Google Colaboratory (Colab), Vector Databases
LLM Expert
2024 - 2025
FluidityIQ, LLC
- Advised on the architectural design and initial implementation of a hybrid search system utilizing vector and keyword retrieval across a 100+ million document global patent database.
- Delivered executive-level technical strategy directing the selection and deployment of core ML components, including embedding models, a multi-stage re-ranking pipeline, hybrid search, and Agentic RAG architecture.
- Worked on R&D and prototyping of a novel patent embedding methodology based on patent claims, which enhances semantic search precision for complex technical queries.
Technologies: Machine Learning, Natural Language Processing (NLP), Large Language Models (LLMs), Artificial Intelligence (AI), Data Science, Python, Search, Agentic AI, Vector Data, Retrieval-augmented Generation (RAG), AI Agents, Prompt Engineering
AI Developer
2024 - 2025
Law Offices of David A. Bressman, CO. LPA
- Designed and developed an AI assistant to support legal scenarios, including back-end algorithms and front-end user interface.
- Fine-tuned a language vision model to extract diagrams from accident reports.
- Developed algorithms for workflows related to handling medical reports and medical expenses in legal scenarios.
Technologies: Artificial Intelligence (AI), Python, Large Language Models (LLMs), Retrieval-augmented Generation (RAG), Natural Language Processing (NLP), OpenAI, Search, Fine-tuning, Solution Architecture, OpenAI API, Agentic AI, AI Agents, Prompt Engineering, Vector Data
AI Architect
2024 - 2025
Solvent.Life Ltd
- Designed and developed a large language model (LLM) pipeline for an AI assistant in financial markets.
- Implemented advanced charting for financial market data, integrated with the AI assistant.
- Defined the architecture for the application and cloud deployment in Google. Integrated payments with Stripe.
Technologies: Python, Artificial Intelligence (AI), Natural Language Processing (NLP), Architecture, Large Language Models (LLMs), Data Lakes, Data Visualization, Dashboards, Llama 2, Finance, Yahoo! Finance, LangChain, OpenAI, Google Cloud, FastAPI, Supabase, Stripe, Stripe API, Solution Architecture, Prompt Engineering, OpenAI API, AI Agents, Agentic AI, Vector Data
LLM and Generative AI Expert
2024 - 2024
Influencer Marketing Company
- Designed an AI content generation system tailored for the influencer content and marketing scenario.
- Implemented automatic optimization of LLM prompts to maximize content quality.
- Developed and deployed a UI to streamline human interaction with the AI system.
Technologies: Natural Language Processing (NLP), Artificial Intelligence (AI), Machine Learning, Large Language Models (LLMs), Generative Artificial Intelligence (GenAI), Python, Data Science, Agentic AI, AI Agents, Prompt Engineering
AI/ML Engineer (via Toptal)
2023 - 2024
Mirzacles
- Fine-tuned several models with client data using QLoRA (Llama 2 70B, Mistral 7B, Mixtral 8x7B, and Yi 34B).
- Developed SFT instructions and trained a judge model for quality evaluation.
- Developed a Gradio prototype app for model testing and evaluation.
- Deployed models with vLLM and FastAPI. Optimized model deployment with Activation-aware Weight Quantization (AWQ).
Technologies: Artificial Intelligence (AI), Machine Learning, Python, Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), Deep Learning, Supervised Learning, Unsupervised Learning, PyTorch, Data Science, Gradio, Large Language Models (LLMs), Llama 2, Machine Learning Operations (MLOps), Google Cloud Platform (GCP), FastAPI, Product Management, LangChain, Generative Pre-trained Transformer 3 (GPT-3), OpenAI, Text Generation, Serverless, Docker, REST, Generative Artificial Intelligence (GenAI), Cloud, Model Tuning, Fine-tuning, Prompt Engineering
Machine Learning Engineer
2023 - 2023
Hum Nutrition Inc
- Developed a conversational AI agent for data analysis from text descriptions. The application has access to information on the internal databases and data structures for retrieval augmented generation and is able to generate and execute SQL queries.
- Developed an AI agent for customer support, using a generative LLM, retrieval augmented generation, and a sequential chain of prompts that implements the theory-of-mind algorithm.
- Deployed the AI agents in Google Cloud and Vercel platforms.
Technologies: Machine Learning, Python, Llama 2, Generative Pre-trained Transformers (GPT), OpenAI GPT-4 API, LangChain, Chatbots, AIOps, Google Cloud Platform (GCP), Supabase, Vercel, Retrieval-augmented Generation (RAG), A/B Testing, Data Scientist, Statistics, Algorithms, Analytics, OpenAI, Text Generation, Serverless, Docker, Predictive Analytics, Statistical Modeling, Data Modeling, SciPy, Data Analysis, REST, Google BigQuery, Generative Artificial Intelligence (GenAI), Cloud, Agentic AI, AI Agents, Prompt Engineering, Vector Data
NLP and Machine Learning Developer (Freelance)
2018 - 2023
Talkmap (formerly Discourse.ai)
- Developed machine learning models for natural language understanding, concept discovery, incremental discovery, and conversational analytics.
- Implemented models for dialog flow identification and classification.
- Built unsupervised models for intent discovery and natural language generation.
- Designed and built the AI platform for incremental discovery of new customer intents in call center conversation data.
Technologies: Continuous Deployment, Continuous Integration (CI), Rasa.ai, GitLab CI/CD, Apache Kafka, Apache Airflow, Data Pipelines, Docker, Big Data, PostgreSQL, Data Science, Machine Learning, PyTorch, Artificial Intelligence (AI), Natural Language Processing (NLP), Generative Pre-trained Transformers (GPT), Natural Language Understanding (NLU), Computational Linguistics, Pytest, SpaCy, Data Engineering, Kubernetes, Amazon Web Services (AWS), Scrum, Scikit-learn, Google Cloud Platform (GCP), Jupyter, Deep Learning, Chatbots, Statistical Data Analysis, NumPy, Python, Pandas, IBM Watson, React, TypeScript, Generative Pre-trained Transformer 3 (GPT-3), OpenAI, HTMX, ChatGPT, Technical Leadership, Software Architecture, FastAPI, Machine Learning Operations (MLOps), Language Models, Large Language Models (LLMs), Haystack, Supervised Learning, Unsupervised Learning, Retrieval-augmented Generation (RAG), A/B Testing, Data Scientist, Statistics, Algorithms, Analytics, Software as a Service (SaaS), Text Generation, Predictive Analytics, SciPy, Data Analysis, REST, Generative Artificial Intelligence (GenAI), Cloud, Model Tuning, Clustering, Hierarchical Clustering, DBSCAN, Text Mining, Data Mining, Fine-tuning, Search, Prompt Engineering, Vector Data
Freelance Chatbot Developer
2018 - 2019
Bigger Brains
- Developed a natural language understanding model for an educational chatbot.
- Integrated the chatbot with Slack, Facebook Messenger, and Microsoft Teams platforms.
- Implemented the chatbot dialog engine using a Rasa machine learning system.
Technologies: Machine Learning, Artificial Intelligence (AI), Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), Natural Language Understanding (NLU), Computational Linguistics, SpaCy, Scikit-learn, Deep Learning, Chatbots, NumPy, Python, Pandas, Django, Rasa.ai, TensorFlow, Software Architecture, Full-stack, Azure, Software as a Service (SaaS), Web Development, Text Generation, Data Analysis, Cloud
NLP/ML Developer
2017 - 2018
Cargo Chief
- Developed a system for data extraction from natural language sources.
- Implemented both rule-based extractions and machine learning for natural language processing.
- Developed tools for data curation and data preprocessing.
Technologies: Amazon Web Services (AWS), Optical Character Recognition (OCR), Data Pipelines, Image Processing, MySQL, Data Science, Machine Learning, PyTorch, Artificial Intelligence (AI), Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), Natural Language Understanding (NLU), Computational Linguistics, SpaCy, Data Engineering, LightGBM, Scrum, Flask, Scikit-learn, GitHub, Jupyter, SQL, Deep Learning, Statistical Data Analysis, NumPy, Python, Pandas, Software Architecture, Supervised Machine Learning, Data Scientist, Software as a Service (SaaS), Data Modeling, Data Analysis, Cloud
Freelance Data Scientist
2017 - 2017
HumNutrition
- Implemented a product recommendation system that was also integrated with the webshop.
- Built a predictive model for the churning of subscription customers.
- Performed a clustering analysis of the customer space to extract insights for marketing.
- Developed REST web services to integrate the models into the main business application.
Technologies: ETL, Data Pipelines, Docker, MySQL, Data Science, Machine Learning, PyTorch, Forecasting, SpaCy, Data Engineering, Recommendation Systems, XGBoost, LightGBM, Scikit-learn, GitHub, Jupyter, SQL, Deep Learning, Web Scraping, Statistical Data Analysis, NumPy, Python, Pandas, TensorFlow, Software Architecture, Full-stack, Data Scientist, Statistics, Algorithms, Analytics, Software as a Service (SaaS), Fraud Prevention, Data Analytics, Statistical Modeling, Data Modeling, Tableau, SciPy, Data Analysis, REST, Google BigQuery, Matplotlib
Freelance Data Analyst
2014 - 2017
Nespresso
- Worked on the blueprint design and implementation of a new retail information system with responsibilities on customer service and supply chain flows.
- Developed web services for eCommerce integration with a supply chain covering stock, availability, and sales documents.
- Implemented an outlier detection system for alerts on customer service data entry.
- Developed an extensive set of automated unit tests for end-to-end supply chain flows.
- Analyzed data log for bottleneck identification and performance improvement.
Technologies: ETL, Data Pipelines, Data Science, Machine Learning, Forecasting, Data Engineering, XGBoost, LightGBM, Scikit-learn, Jupyter, SQL, Statistical Data Analysis, NumPy, Python, Pandas, R, SAP, Data Mining, Big Data, Data Scientist, Statistics, Algorithms, Analytics, Data Analytics, Predictive Analytics, Statistical Modeling, Data Modeling, Test-driven Development (TDD), SciPy, Data Analysis, Matplotlib
Freelance Data Analyst and Developer
2013 - 2013
Syncronic
- Worked as a core developer of a new SaaS platform for sales forecasting.
- Created the general architecture design, forecasting model, back-end development, REST API, and integration with the ERP system.
- Developed a forecasting model by combining ARIMA and ETS with a machine-learning gradient-boosting model.
Technologies: ETL, Data Science, Machine Learning, Forecasting, Scikit-learn, Statistical Data Analysis, NumPy, Python, Pandas, Time Series, Technical Leadership, Software Architecture, Full-stack, Statistics, Algorithms, Analytics, Software as a Service (SaaS), Data Analytics, Statistical Modeling, Data Modeling, SciPy, Data Analysis
Data Analyst (Contract)
2011 - 2013
Novozymes
- Worked on the business blueprint and implementation of a new SAP supply chain solution.
- Designed and developed new production planning heuristics.
- Developed automated end-to-end tests for the business processes.
Technologies: ETL, Data Science, Machine Learning, Scikit-learn, Jupyter, SQL, Statistical Data Analysis, NumPy, Python, Pandas, SAP, Data Pipelines, Data Scientist, Algorithms, Analytics, Data Analytics, Predictive Analytics, Data Modeling, Data Analysis
Web Developer (Contract)
2011 - 2011
Unilever
- Enhanced the Python back-end of Digpedia.net—a multimedia marketing internal web application.
- Developed, in Django, a system for multimedia management with fine-grained access control.
- Implemented in the Django application a system for temporary access to resources.
Technologies: Python, PostgreSQL, Django, HTML, CSS, JavaScript, Technical Leadership, Full-stack, Web Development, jQuery
Data Analyst | Developer (Contract)
2008 - 2011
Tetra Pak
- Designed and developed a block-planning extension to the production planning software.
- Collaborated on the design of the new available-to-promise solution and product-allocation solution.
- Designed and implemented enhancements to the SAP CIF interface to support a special VMI scenario.
- Supported and worked on solution enhancements for rollouts in multiple plants.
Technologies: Data Science, SQL, Statistical Data Analysis, Python, SAP
Data Analyst (Contract)
2009 - 2010
Sony Mobile
- Implemented a prototype for the available-to-promise system using rules-based functionality for the project proof of concept.
- Designed and implemented an SAP global available-to-promise system.
- Designed and built the technical specifications for enhancements in the APO interface, backorder processing, and special reports.
Technologies: Data Science, Python, SAP, Data Analytics
Data Analyst (Contract)
2008 - 2010
Nestlé
- Worked on the data analysis for performance improvement and issue resolution.
- Designed and implemented an algorithm to support products with shelf-life constraints in sales stock allocation.
- Enhanced the available-to-promise process in the supply chain management system.
- Designed and developed a simplified user interface for quotas and allocation maintenance. Used dynamic source code generation.
Technologies: Data Science, ABAP, SQL, Python, SAP
SCM Consultant
2007 - 2008
SAP
- Acted as a member of the MaxAttention team and was sent on a mission to Sappi paper company in South Africa to identify and propose solutions for the critical issues in the SAP system.
- Designed the detailed plan for implementation of Available to Promise in Philips Consumer Electronics.
- Analyzed and proposed a solution for a timezones issue in ATP scheduling in the SAP system of Clariant.
Technologies: SAP
Data Analyst (Contract)
2007 - 2007
Philips Consumer Electronics
- Collaborated on the business blueprint design for supply chain quota allocation as a technical expert.
- Implemented a prototype with the algorithm identified in the blueprint.
- Worked on the project SWOT analysis.
Technologies: Data Science, Python, SAP, Data Analytics
Data Analyst (Contract)
2007 - 2007
Johnson & Johnson Pharmaceutical
- Conducted a data analysis in order to troubleshoot system issues and implement improvements.
- Designed and developed enhancements on the sales forecasting and a demand-planning system.
- Implemented enhancements to the production-planning detailed scheduling system.
Technologies: Data Science, SQL, Python, SAP, Data Analytics
Data Analyst | Developer
2001 - 2006
Sonae Industria
- Worked on a new greenfield SAP implementation with a focus on the supply chain system.
- Analyzed data in order to achieve performance and stability improvement.
- Developed a real-time asynchronous interface for a SAP integration with shop-floor control.
- Implemented a web-based performance-monitoring dashboard.
Technologies: Java, Data Warehouse Design, Data Science, ABAP, SQL, Statistical Data Analysis, Python, R, Oracle, SAP, Web Development, Data Analytics
Experience
First Prize Winner of Microsoft Research TextWorld AI Competition
https://www.microsoft.com/en-us/research/blog/first-textworld-problems-the-competition-using-text-based-games-to-advance-capabilities-of-ai-agents/I am the winner of a global competition on reinforcement learning and natural language understanding for building agents capable of solving text-based games.
Kaggle Master
https://www.kaggle.com/pvlimaI achieved the rank of master in a Kaggle machine learning competition platform.
Sales Forecasting Platform
I worked on the design and development of a SaaS platform for sales forecasting. I was also involved in the implementation of a hybrid time series forecasting algorithm using ARIMA, ETS, and gradient boosting.
Production Planning Heuristic
A heuristic algorithm that I designed and developed for the special cyclic planning of fast-moving products. The planning heuristic is based on standard lots planning with additional leveling of production quantities and special scheduling based on product setup values.
Outlier Detection System
I implemented an outlier detection system based on a random forest algorithm to provide alerts for suspicious/wrong data in the customer service flow.
Game with Generative AI
Developed a game with content created by generative AI (poems and game images). The game is featured in the Museum of AI ("Love is like a foot on the beach" by Bot Poets Society 2021; AI-generated poem).
US Patent | US11847422B2
As a result of my R&D work, I was the inventor of the US patent US11847422B2 on conversational AI with the title "System and method for estimation of interlocutor intents and goals in turn-based electronic conversational flow."
Education
1999 - 2007
PhD Degree in Machine learning applied to process engineering
Coimbra University - Coimbra, Portugal
1992 - 1997
Master's Degree in Chemical Engineering
Coimbra University - Coimbra, Portugal
Skills
Libraries/APIs
PyTorch, XGBoost, Scikit-learn, Pandas, Matplotlib, Beautiful Soup, TensorFlow, Keras, SciPy, NumPy, Gradio, OpenAI API, SpaCy, jQuery, React, HTMX, Stripe, Stripe API
Tools
Rasa.ai, Apache Airflow, Jupyter, GitHub, Git, ChatGPT, Haystack, Codex, Claude Code, Sublime Text, GitLab CI/CD, Tableau, PyCharm, IBM Watson, Pytest, Yahoo! Finance
Languages
Python, SQL, ABAP, Java, R, JavaScript, TypeScript, HTML, CSS
Frameworks
LightGBM, Django, Scrapy, Flask
Paradigms
ETL, Continuous Integration (CI), Continuous Deployment, Test-driven Development (TDD), REST, Scrum
Platforms
Docker, Google App Engine, MacOS, Linux, Google Cloud Platform (GCP), Kubernetes, Oracle, Apache Kafka, AWS Lambda, Amazon Web Services (AWS), Vercel, Azure, Blockchain
Storage
Data Pipelines, PostgreSQL, MySQL, ClickHouse, Elasticsearch, PostGIS, MongoDB, Data Lakes, Google Cloud
Other
Data Analytics, Chatbots, Artificial Intelligence (AI), Data Analysis, Statistical Data Analysis, Natural Language Processing (NLP), Data Engineering, Data Science, Machine Learning, SAP, Natural Language Understanding (NLU), Generative Pre-trained Transformer 3 (GPT-3), Generative Pre-trained Transformers (GPT), Text Generation, FastAPI, Language Models, Machine Learning Operations (MLOps), Large Language Models (LLMs), Analytics, Data Scientist, Predictive Analytics, Data Modeling, Generative Artificial Intelligence (GenAI), Clustering, Fine-tuning, Prompt Engineering, AI Agents, Agentic AI, Vector Data, Text Mining, Image Processing, Serverless, Recommendation Systems, lxml, Web Scraping, Deep Learning, Scientific Computing, Computational Linguistics, OpenAI, LangChain, Technical Leadership, Software Architecture, Full-stack, Llama 2, OpenAI GPT-4 API, Supervised Learning, Unsupervised Learning, Retrieval-augmented Generation (RAG), Supervised Machine Learning, Data Mining, Web Development, Fraud Prevention, Software as a Service (SaaS), Algorithms, Statistics, A/B Testing, Statistical Modeling, Outlier Detection, Cloud, Hierarchical Clustering, DBSCAN, Search, Solution Architecture, Forecasting, Time Series, Big Data, Maps, Optical Character Recognition (OCR), Data Warehouse Design, Google BigQuery, Computer Vision, Games, Optimization, AIOps, Supabase, Product Management, Web3, Architecture, Data Visualization, Dashboards, Finance, Model Tuning, Document Processing, AI Pipeline, Pinecone, Google Colaboratory (Colab), Vector Databases
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