Hey there! Nice to see you around here. Whether you're here to explore some cool AI projects, collaborate, or just check out what I’ve been working on, I’m excited to share it with you!
I work at Microsoft in the Global Black Belt AI team as a Principal AI & Application Development Architect. I combine AI, Machine Learning, Data, and software engineering to build agentic GenAI systems that Fortune 500 teams actually use. I turn white-paper theory into production-ready AI—elbows-deep in Generative AI, multi-agent orchestration, and LLMOps. Most days, I’m writing code, shaping strategy, and mentoring the next wave of builders.
- Secure Real-Time Voice Agents: Architecting HIPAA-grade, end-to-end-encrypted speech-to-speech pipelines that keep PHI safe while enabling sub-second agentic conversations.
- Agentic Protocols & Control Planes: Designing open communication standards (MCP, A2A) and mapping them onto Azure AI Foundry’s multi-agent control planes for seamless agent hand-offs.
- Production Agentic Memory Layers: Building hierarchical short-/long-term memory that persist and retrieve context across agent invocations in mission-critical flows.
- LLM-App Evaluation & LLMOps: Creating end-to-end evaluation harnesses to track latency, cost, alignment, and quality of multi-agent GenAI applications before and after release.
- Founder of AI HLS Ignited — Created and led the flagship accelerator program at Microsoft, empowering the HLS industry across live technical webinars and GitHub-first assets. The program is catalyzing real-world GenAI adoption across provider, payor, and life science orgs through hands-on labs and applied use cases.
- Northwestern Adjunct Instructor — Teaching cloud computing and AI to the next wave of leaders in the MSAI graduate program.
- Established and led an MLOps practice, deploying both real-time and batch ML solutions globally at Concentrix consulting services.
- Led multiple algorithms to production-grade ML systems within Fortune 500 enterprises during my consulting stage at Concentrix, from concept to deployment.
- Principal author and designer of the MLOps accelerator framework at Levi's & Co Tech.
- Co-authored, developed, and engineered a set of Python SDKs at Levi's & Co Tech, simplifying ML model integration into software for large DS organizations.
- Designed a state-of-the-art product classification and embedding generation system using advanced Computer Vision algorithms.
- MedIndexer — Transform fragmented clinical documents into schema-driven, high-value knowledge stores using Azure AI Search and AOAI.
- MedEvals — Benchmark LLM/SLM models at scale with multi-metric evaluation, real-world datasets, and LLM-as-a-Judge methods.
- MedAgenticR&D — Enable dynamic multi-agent RAG across structured and unstructured data using Semantic Kernel + Azure AI Agents.
- AutoAuth — Cut prior auth turnaround by 70% with a modular agentic system integrating LLMs, OCR, and policy reasoning.
- MedCoding — Built a multi-agent workflow to extract, validate, and map ICD-10 codes, improving billing accuracy and reducing fraud.
- RTAgent — Deployed real-time voice-to-voice AI pipelines using Azure Speech, OpenAI, and Redis, tailored for healthcare call centers.
- Building AI-Powered Clinical Knowledge Stores with Azure AI Search
- Streamlining Prior Authorization with Generative AI
Learn how AutoAuth reshapes PA workflows with agentic AI. - Key Technical Challenges in GenAI Production
Explore real-world blockers in scaling GenAI solutions. - Effective LLM/SLM Evaluation Framework
A roadmap for scalable, trustworthy LLM evaluation. - Modernize Your Document Management with Azure AI
From OCR to NER—supercharge unstructured data workflows.
I’m always open to collaborate on Open-source challenges, innovative product ideas, and partnerships with bright AI/ML practicioners.
- 💌 Shoot me an email: [email protected]
- 🔗 LinkedIn: Let’s get linked!
- ✍️ My Blog: Pablo's AI Corner