Felipe Araya Olea

• AI Strategy & Leadership: Led the design and delivery of enterprise-scale AI and
predictive analytics solutions supporting strategic finance and executive decision
making across NEOM.
• AI Capex/Opex Classification: Designed and deployed an end-to-end LLM-powered
classification system to automatically categorise purchase orders. integrating several
AI models to perform document curation, RAG-based retrieval, and explainable AI
outputs within SAP workflows
• AI Insurance: Built entity extraction pipelines for insurance policies and
endorsements, accurately extracting 30+ structured fields using OCR, RAG, symbolic
AI, validation rules, and expert review, achieving ~90% fully error-free documents.
• Executive First AI: Led the development of a data & forecasting chatbot for NEOM
executives to enable them to fully understand the entire Neoms data and forecasting
numbers, including population, tourism, employment, energy, HR, and healthcare.
• Tech Stack: Python, SQL, Azure, Microsoft Fabric, SAP, AI Search, RAG, NLP, LLM,
SQL, LangChain, LangGraph, LlamaIndex, AutoGen, LLMOps, Data Governance,
Docker, Cloud Functions.

• Algorithms from Scratch: Co-develop and taught an advanced program in AI &
Data Science, which included the implementation of machine learning, deep learning
and Agentic AI theories from scratch.
• Technical Topics: Designed the content of several topics such as Generative AI,
Agentic AI, Unsupervised Learning, and detailed intricacies of tree-based algorithms
(Gradient Boosting and XGBoost).
• Tech Stack: Python, SQL, Git, LLM, RAG, Agentic AI, Guardrails, Generative AI.

Just Eat Takeaway is a leading global online food delivery marketplace, connecting millions of consumers with restaurants and couriers.

• Logistics & Courier Pay Models: Working in the logistics domain, designing and enhancing models that optimize courier pay while balancing operational costs for the business.
• Optimization: Applied advanced optimisation algorithms to simulate different pay strategies, ensuring fair compensation for couriers while maintaining business profitability.
• Simulations & Modeling: Built robust simulations to test different courier pay policies using historical data and future demand scenarios.
• Cloud Platforms: Developed and deployed scalable solutions using GCloud (Vertex AI, Vertex Pipelines) and AWS to support model training, monitoring, and serving.
• Containerization & Pipelines: Utilized Docker for containerized workflows and automated data pipelines for production-level delivery.

• AI Solution: Developed a clinician AI platform used across NHS trusts to convert
clinician-patient conversations into structured, compliant clinical notes that can be
used to help diagnostics and patient medical record keeping, registered as a Class I
medical device.
• End-to-end: Led the design, creation, and delivery of ClearNotes from end-to-end,
including problem definition and stakeholder alignment through prioritization.
• Innovation: I solely created a proprietary multi-layer AI quality and validation
framework that will verify and validate multiple key points; audio quality, transcription
accuracy, summarization quality, and clinical safety, using a multi-stage LLM.
• Tech Stack: Python, SQL, Azure, AWS, OpenAI (Azure-hosted), RAG, LLM, Agentic
AI, Deepgram, Speechmatics, Whisper, Amazon Transcribe, Vector Databases,
Advanced Analytics, Data Governance.

• Purpose: E-commerce platform designed to assist customers with product insights and
market analysis to improve their market strategies on Bol.com and Amazon
• Time Series and ML: Created a robust ML (XGBoost) model capable of predicting
sales of thousands of products across different category levels.
• API: Created an API to handle product queries, giving insights and sales forecasting
using Flask and serverless frameworks AWS Lambda.
• ETL: Designed ETL pipelines using Airflow, AWS, async solutions, and Spark to
orchestrate and automate the feature engineering process.
• Tech Stack: Python, SQL, Airflow, PyTorch, AWS, ETL, Git, Digital Ocean,
Snowflake.

• Purpose: Homemove is a platform offering various moving-related services, including
surveys, removals, and mortgages, all integrated within a single app.
• AI tools: Led the development an LLM-powered negotiation tool, which allows users to
obtain quotes and negotiate prices in real time, add customers to our CRM system
automatically, and alert the sales team upon successful negotiation.
• Data Transformation and ETL: Led the data transformation initiative, leveraging
Snowflake, dbt, AWS, and Airflow to automate data ingestion and transformations.
• Likelihood to sell a property: Developed an ML model aim at reducing marketing costs
and improving targeting by identifying home movers in the next 6 months
• Tech Stack: Airflow, Snowflake, dbt, SQL, Python, AWS, Git, PyTorch.

Engineer
• Purpose: Property insight company that specialises in providing clients with targeted
data at a household level to focus on their resources and delivery of their services.
• AI Tools: Built LLM model assistants for code documentation and feature extraction
from property listing descriptions through RAG models connected with Confluence
and Zoopla API.
• Machine Learning: Created two major models that became the champion models for
the company; Propensity to move home and propensity to list your property using
XGBoost and LightGBM.

• Machine Learning: Created and worked on several models such as cashflow models,
early redemption models, default models, pre-payment models, and conversion
models using time-series, classification, and survival analysis.
• Tech Stack: Python, R, XGBoost, ARIMA, Random Forest, Survival Analysis, Azure
DevOps.

• Machine Learning: Worked on several models such as lifetime value, sales
forecasting, and customer segmentation
• Tech Stack: Python, logistic regression, Random Forest, deep learning, neural
networks.