Coming Soon

What I'm building right now.

Guided Visualization

In progress

An app that writes an original guided visualization every session, choosing from six generation templates — from calming pre-competition nerves to winding down for sleep. Nothing is pre-recorded; each one is generated fresh, built to feel like a memory, not a script.

Text pipeline is built and wired to Claude, with six session templates each validated against a hand-written exemplar. Audio narration is next.

Claude API (Anthropic SDK)Prompt engineeringAudio-first UX

Right Whale Classifier

In progress

A planned computer-vision model to identify individual North Atlantic right whales, using a photo-based dataset.

Researchers already track right whales individually by the unique callosity patterns on their heads. The idea: fine-tune a CNN on that same identification task and see how close a from-scratch model gets.

PythonComputer visionCNN / transfer learningImage classification

Wildlife Drone Program

In progress

Building both the aircraft and the autonomy stack behind a deliberately homegrown goal: spotting the coyotes that move through the neighborhood, from the air.

A hardware build and a software program at the same time. Three airframes — an X500 V2 as the proving platform, a Fighter 2430 VTOL for long-endurance mapping, and an RJX1300H hexacopter, the largest of the three, for close-range multi-sensor work — assembled and bench-tested rather than bought turnkey, with a Jetson companion computer carrying the perception load. The software runs PX4 and MAVSDK under one hard rule: flight-critical authority stays on the autopilot, covering geofence, return-to-home, battery and link-loss failsafes, while mission logic and inference stay on the companion computer. The current milestone is simulation-only — an arm/takeoff/loiter/land smoke mission against PX4 SITL, a config layer written to import no MAVSDK at all so it unit-tests in CI with no simulator or vehicle present, and an address check that refuses anything but loopback, so the smoke mission structurally cannot be pointed at a real aircraft.

GitHub ↗
PX4 / MAVSDKPythonROS 2Jetson companion computeComputer visionAirframe buildGitHub Actions CI

AI Education Roadmap

37 courses and certifications I plan to work through, in the order I intend to take them — fundamentals first, then hands-on building, then depth. Updated as I finish them.

  1. 1

    Currently working through

    Internal EY training, running alongside everything below.

    • EY Sustainability Courses
    • EY Technology Courses
  2. 2

    Cloud & data fundamentals

    The vocabulary layer — what the services are, and where AI and data sit in a cloud stack.

    • Microsoft Azure Fundamentals (AZ-900)
    • Microsoft Azure AI Fundamentals (AI-901)
    • Microsoft Azure Data Fundamentals (DP-900)
  3. 3

    Building with LLM APIs & agents

    Free, fast, and closest to what I'm already building — these feed the projects on this site directly.

    • Anthropic Academy — Building with the Claude API
    • Anthropic Academy — Claude Code 101
    • Anthropic Academy — Claude Code in Action
    • Anthropic Academy — Introduction to Model Context Protocol
    • Anthropic Academy — Advanced MCP
    • Anthropic Academy — Agent Skills
    • Anthropic Academy — Subagents
    • OpenAI Academy — AI Foundations
    • OpenAI Academy — Agents and Workflows
    • Hugging Face — LLM Course
    • Hugging Face — AI Agents Course
    • DeepLearning.AI — Multi AI Agent Systems
    • DataCamp — Developing AI Applications with the OpenAI & Anthropic APIs
  4. 4

    Data platform & BI

    The stack client work actually runs on. My projects prove Python; this proves the rest of the pipeline.

    • Microsoft Certified: Power BI Data Analyst Associate (PL-300)
    • Microsoft Certified: Fabric Data Engineer Associate (DP-700)
  5. 5

    Machine learning depth

    The part behind the Data Science tab — so the CNN and clustering work rests on theory, not just tutorials.

    • DeepLearning.AI — Machine Learning Specialization
    • DeepLearning.AI — Deep Learning Specialization
    • Kaggle — Feature Engineering and Intro to Deep Learning
    • Microsoft Certified: MLOps Engineer Associate (AI-300)
  6. 6

    Applied Azure AI

    Hands-on assessments rather than multiple choice — you build the thing to pass.

    • Microsoft Applied Skills — Develop generative AI apps using Azure AI Foundry
    • Microsoft Applied Skills — Get started developing agents in Microsoft Foundry
    • Microsoft Certified: Azure AI Apps and Agents Developer Associate
  7. 7

    Vendor specializations

    Depth in the model and data-platform layer.

    • NVIDIA-Certified Associate — Generative AI LLMs (NCA-GENL)
    • NVIDIA-Certified Associate — Generative AI Multimodal (NCA-GENM)
    • Databricks Certified Generative AI Engineer Associate
  8. 8

    Risk, sustainability & governance

    Where the accounting background actually compounds — the AI work is more useful if I can speak climate disclosure and model risk too.

    • GARP — Sustainability and Climate Risk (SCR)
    • IFRS Foundation — Fundamentals of Sustainability Accounting (FSA) Credential
    • IAPP — AI Governance Professional (AIGP)
  9. 9

    Stretch goals

    Breadth once the core path is done. NCP-AAI expects a year or two of production agentic work, so it's genuinely last.

    • Google Cloud — Generative AI Leader
    • AWS Certified AI Practitioner (AIF-C01)
    • NVIDIA Certified Professional — Agentic AI (NCP-AAI)
  10. 10

    Outside the AI track

    A side project rather than part of the path above — though a drone and the computer-vision work would pair well eventually.

    • FAA Part 107 — Remote Pilot Certificate (small UAS)