Diego Marinho de Oliveira

AI/ML Leader | Agentic AI, AI Search, and Recommendations | Production AI at Scale

Melbourne, Victoria, Australia · dmarinho.ai@gmail.com

Executive Summary

• AI/ML leader with 10+ years delivering Agentic AI, AI search, and recommender systems across Oceania, Asia, and South America.
• Designed, led, and deployed production-grade AI agents and GenAI solutions at scale with continuous innovation.
• Proven business impact through 150+ online experiments and multiple customer-facing AI launches.
• Knowledge-driven and dedicated: top student, ANN/NLP research, teaching assistant, and delivery across startups and enterprise companies.

Leadership and Impact

• Drove multi-market recommender transformations at SEEK with 5-10x gains and a 12x record in an Asia market.
• Led AI initiatives across Agentic AI, AI search, multimodal CV, and recommendations.
• Built REA’s first GenAI property search in under 30 days and shipped production AI agents for direct user conversations about properties.
• Delivered real-time property photo semantic attribute extraction using multimodal LLM/transformers (45+ attributes).
• Shipped multilingual personalized AI search for HK (JobsDB) with +14% improvement on a core metric.
• Delivered multiple other production AI launches beyond the flagship projects listed below.

Operating at Scale

• Ran 150+ online experiments and A/B tests with positive impact on primary metrics; record uplift in an Asia market.
• Delivered real-time recommendations for millions of users across homepage, email, and push notifications.
• Owned cost and performance targets across AWS/GCP; optimized LLM usage with caching and token-level cost tracking.
• Designed systems with offline evaluation, online testing, and load tests to meet SLAs (recs ~500ms, GenAI 1-2s).
• Implemented guardrails, rate limiting, and LLM-as-judge monitoring; completed a security pen test.
• Defined OKRs with executive stakeholders and partnered across product, engineering, and leadership.

Signature Projects

• GenAI Property Search (REA): Built and launched in under 30 days; production AI search for property discovery. Media
• Agentic AI Platform (REA): Designed and deployed AI agents at scale with cost controls, safety, and observability. Media
• Computer Vision Attribute Extraction (REA): Property photo semantic attribute extraction using multimodal LLM/transformers, 45+ attributes in production.
• Unified Recommender Platform (SEEK): 8+ markets, 5-10x gains, and a 12x record in an Asia market.
• Personalized AI Search (SEEK JobsDB): +14% improvement on a core metric in HK.

Professional Experience

Senior Engineering Manager / Principal AI — Machine Learning, REA Group (Melbourne, Australia)

April 2024 — Present

• Lead DS/MLE teams delivering production Agentic AI, AI search, and personalization (12 people, up to 3 teams).
• Built a GenAI property search engine for the RED brand end-to-end in under one month.
• Launched the 1st Conversational AI agent for the RED brand and met SLA 1-2s.
• Designed and deployed AI agent systems to production at scale.
• Delivered real-time CV attribute extraction and multimodal inference in production.
• Implemented GenAI observability for real-time monitoring, evaluation, and tracing.
• Managed GenAI costs (input/output, caching) and met SLAs (1-2s per request).
• Applied guardrails, rate limiting, and LLM-as-judge monitoring; completed a security pen test.

AI / Data Scientist Manager; Senior Data Scientist; Data Scientist, SEEK (Melbourne, Australia)

March 2017 — April 2024

• Managed a global DS team of ~10 and led hiring processes across multiple AI teams.
• Led recommender system transformation with 5-10x improvements and a 12x record in an Asia market.
• Delivered a unified recommender platform across 8+ markets in Asia and ANZ.
• Launched personalized AI search for HK (JobsDB) with +14% main-metric improvement.
• Ran 150+ online experiments and A/B tests with positive impact on main metrics.
• Served millions of users with real-time recommendations across homepage, email, and push notifications.
• Designed systems with offline evaluation, online testing, and load tests to meet SLAs (~500ms for recs).

Lead Data Scientist, Catho Online (Sao Paulo, Brazil)

July 2015 — February 2017

• Built candidate recommendation systems with >115% uplift in main metrics.
• Delivered hybrid recommenders combining collaborative filtering and content-based methods.

Education

• M.Sc. Computer Science, Universidade Federal de Minas Gerais (2010-2012) Focus: NLP/NER; FS-NER research (see Publications). Top student, 90%+ average.
• B.Sc. Computer Science, Pontificia Universidade Catolica de Minas Gerais (2006-2009) Student Medal for Excellence; 90%+ average. Research in ANN; tutoring; internships.

Awards (Selected)

• Winner, REA Hackdays 2024 — Hack it Forward Award for Community, Social & Environmental Good (2024)
• Top performer, Databricks Gen-AI Australia Cup (2023)
• Winner, SEEK International Volponi Award (2017)
• Student Medal for Excellence, B.Sc. Computer Science (2009)

Publications (Selected)

• Imbalance Data Sparsity as Source of Unfair Bias in Collaborative Filtering (RecSys 2022)
• Offline Evaluation Standards for Recommender Systems (RecSys 2021)
• FS-NER: Filter-Stream Named Entity Recognition on Twitter Data (WWW Workshop 2013)
• Applying Neural Networks to Determine the Socio-Environmental Factors Responsible for Potable Water Consumption (SMC 2010)
• The Usage of Artificial Neural Networks in the Classification and Forecast of Potable Water Consumption (IJCNN 2009)
• New Perspectives in the Context of Computer Architecture – Educational Software for Parallel Computation (WEAC 2009)

Skills

• Leadership: org design, hiring, mentoring, cross-functional strategy.
• AI/ML Systems: ranking, retrieval, AI search, semantic search, sequential recommenders, online experimentation, A/B testing, learning-to-rank.
• GenAI/LLMs: agentic AI, AI agents, RAG, embeddings, structured outputs, fine-tuning (QLoRA), evaluation.
• ML Techniques: transformers, embeddings, matrix factorization, clustering, RL, offline evaluation.
• Cloud Platforms: GCP, Vertex AI, AWS, BigQuery.
• Tools: PyTorch, TensorFlow, scikit-learn, Pandas, Spark, FastAPI.

Core Expertise

AI/ML leadership, Agentic AI, AI search, ranking systems, retrieval, recommender systems, GenAI, LLMs, RAG, MLOps, observability, safety, guardrails.