About
Building AI systems that work in the real world
A portfolio of AI systems, experiments, and prototype builds — grounded in two decades of real-time engineering across broadcast, streaming, and analytics platforms.
01
What I Do
End-to-end AI systems
I build end-to-end AI systems with a strong bias toward working software, clear evaluation loops, and deployment-ready architecture. The work sits at the intersection of machine learning, software engineering, and product design.
Most projects fall into three areas: agentic AI systems for structured workflows, voice AI applications for natural interaction, and learning-oriented tools with measurable feedback. Each one is treated as a practical system, not just a prompt demo.
02
Foundation
Twenty years in real-time systems
Before focusing on LLM-powered products, I spent over 20 years building real-time media and AI systems across broadcast, streaming, and analytics platforms. That background shapes how I approach AI today: clear architecture, operational reliability, and systems that perform under real usage.
Real-time streaming and media pipelines
Low-latency interaction systems
Video analytics and AI-assisted workflows
Distributed processing and service orchestration
Production deployment on cloud infrastructure
03
Expertise
Technical depth
AI & Machine Learning
Large Language Models (GPT, Claude, Llama)
Agentic AI & multi-agent systems
Retrieval-Augmented Generation (RAG)
Speech recognition & NLP
Prompt engineering & fine-tuning
Model evaluation & optimization
Software Engineering
Python, TypeScript, React
FastAPI, Node.js, Next.js
PostgreSQL, MongoDB, Redis
Docker, AWS, GCP cloud infrastructure
RESTful & GraphQL APIs
CI/CD & DevOps practices
04
Approach
How I work
I start from the interaction design and system boundaries, then work backward into architecture, orchestration, and evaluation. That keeps the build grounded in what the user actually experiences.
Most ideas are tested first as small prototypes or reference implementations — easier to validate feasibility, identify hidden constraints, and avoid overbuilding too early.
With teams, I prefer short feedback loops, visible progress, and direct tradeoff discussions. The goal is always a system that survives real usage, not just a polished demo.
Beyond the code: when I’m not building AI systems, I’m behind a camera. Photography is a different kind of problem-solving — composition, light, and story — and it keeps the creative instincts sharp.
Open to the right conversations
If you’re working on something adjacent — product direction, prototype builds, or AI system architecture — send a note with the problem space and current constraints.
Start a Conversation