Lorenzo Orozco

Founder · AI Product Studio

• Founded 2Sunrise in Mexico City. Build AI products for 5+ clients, from deciding what is worth building to running the software after launch.
• WUF: built a WhatsApp pet-care agent that picks up where the conversation left off, follows up, and takes actions. Uses LangGraph, Node.js, and the WhatsApp Business API.
• 10-5-3: built an agent that reads monthly reports from 40+ branches, compares them with last month, checks compliance, and prepares a review for the commercial director.
• Created our engineering framework for client work. Every ticket needs agreed acceptance criteria and evidence that the work is done, supported by CI checks, git hooks, skills, and automations.
• Studio projects also include a purchase flow with Lockton for Costco members and an accounting-data pilot for AGSA.

Software Engineer · Platform Engineering / AI4IT

• Lead a team of five in Platform Engineering / AI4IT. Work with business teams to define what needs building, turn it into a technical plan, and help the team deliver.
• Led 10 vendor developers in building a shared engineering framework and developer toolbox for L’Oréal’s order management systems (OMS) teams.
• Built and supported three Python inventory services through QA for SalonCentric. Validated 7,279 SKUs across 7 distribution centers, with 15-minute stock updates and a nightly resync, using SAP, Kafka, Cosmos DB, and Azure.
• Corrected inventory rules that made unavailable stock appear sellable. A QA check of 150 SKUs found 500 stale warehouse positions; added zero-fill updates and fixed shared-barcode totals.
• Built the commission job and on-demand Flask API so finance can trace who gets credit and why. Used territory history and SFTP delivery to Xactly and finance reporting; contributed to Kafka ingestion.
• Designed and built QA Analyst and QA Tester agents plus four skills that turn story requirements into Gherkin tests and evidence workbooks. Piloted the delivery framework on five inventory stories.
• Used parity tests on real QA data to check inventory refactors. The job has 400 tests and 98.8% coverage; commission allocation has about 300 tests and 88.9% coverage.

AI Engineer

• Built retrieval for legal documents so answers can be traced to their source. Combined OCR and LangGraph with hybrid search using embeddings, BM25, and FAISS. Evaluations showed 30% higher recall and 92% question-extraction precision.
• Built document-ingestion and ETL pipelines for 100K+ pages per month, with Azure deployment and prompt evaluation.

Machine Learning Engineer

• Took computer vision from labelled images to YOLOv8 running on Raspberry Pi 5. Used ONNX Runtime for inference and Flask for APIs.
• Built EVA with sensor dashboards and search over technical manuals. Also integrated Copilot Studio and CRM APIs to handle customer queries.

Researcher · Physics-Informed Neural Networks

• Built physics-informed models for solar absorption. Reached MAE 0.00065 and R² 0.92 in research evaluation, compared 1,710 curves, and built a Flask API to try the predictions.

Machine Learning Research

• Trained a model to predict CO₂ transfer coefficients, reaching R² = 0.99. Built a tuning loop to find architectures that met the MAE targets.