Work

Projects and labs

Flagships first: things with a live demo or a public repo. Coursework and walkthroughs sit below so the proof is easier to find.

Flagships

Shipped and hands-on

Agents FastAPI Live

Predictive maintenance multi-agent system

A multi-agent system that monitors industrial equipment, detects anomalies, estimates failures, and surfaces maintenance recommendations.

Sensor monitor uses an Isolation Forest across temperature, vibration, pressure, current, and rotation speed.
Diagnostic agent uses a Random Forest classifier for failure prediction, root cause notes, and a 24–120 hour time-to-failure range.
Dashboard supports live updates, health status, alerts, and CSV or Excel uploads.

Stack: Python, FastAPI, scikit-learn, Plotly, WebSocket, Pandas, Docker, Render.

Open live dashboard

Edge Jetson

Edge AI on NVIDIA Jetson

Hands-on work with a Jetson Orin Nano developer kit to run generative tasks locally, plus NVIDIA’s guide to LLMs, vision-language models, and robotics foundation models on Jetson.

Stack: Linux, JetPack, Docker, TensorRT, CUDA, PyTorch.

Jetson AI Lab NVIDIA guide

Hugging Face Repo

Image-to-audio story

An app that takes an image and turns it into a spoken story: image-to-text, a language model, then text-to-speech, using a Hugging Face token.

Stack: Hugging Face, Python.

View code

Labs and coursework

Practice projects

These are learning builds and platform walkthroughs. Titles are cleaned up; they are not client engagements.

Applications

ChatGPT API assistant

A personal assistant for schedules and quick lookup, scoped in Python with the ChatGPT API and a Gradio interface. The prompt work targeted personalized learning experiences on GPT-3.5 or GPT-4.

Stack: Python, OpenAI API, Gradio.

Generative apps with Gradio

Interfaces for Hugging Face models covering summarization, image-to-text, text-to-image, and language models.

Stack: Hugging Face, Gradio, Python, Transformers.

Data assistant with Streamlit, LangChain, and OpenAI

Environment setup, package imports, a Streamlit interface, and a language-model connection so a data script becomes an interactive app.

This website

The personal site itself: HTML, CSS, and JavaScript used to document projects and the AI stack. An earlier version started as a GitHub Pages exercise.

Cloud and MLOps

MLOps pipeline on Google Cloud

Adapted an open-source pipeline that applies supervised fine-tuning so a language model answers user questions more usefully. Covered data and model versioning, and preprocessing large datasets in a warehouse.

Stack: Google Cloud, Vertex AI, Kubeflow, Kubernetes.

ServiceNow agent exposure

Coursework on defining a problem, training a model, and placing it in a UI, plus exposure to ServiceNow modules (ITSM, HAM Pro, Now Assist, ITOM) and conversational agent patterns.

Stack: ServiceNow, Python, JavaScript, TensorFlow, Docker, Kubernetes, Flask.

Vision and ML foundations

Computer vision on LandingLens and Azure AI

Practice with large vision models on LandingLens and Azure AI Studio: training on image data and deploying a vision workflow.

AI development with Python

Environment setup, Pandas exploration, model training, Hugging Face hubs, and APIs for language, speech, and image generation.

Reference, not a personal product

Vercel AI SDK

Notes on Vercel’s open-source SDK for streaming AI apps in JavaScript and TypeScript: one API across providers, chat UIs, tools, and structured output. This is a toolkit reference, not a product I shipped.

AI SDK cookbook