AI Platform Engineer & Cloud Native Architect
Engineering leader in distributed systems with a core philosophy that resilient, scalable systems are built by teams with a deep sense of shared ownership, fostered through testable code, clear documentation, and transparent collaboration. Experience spans from globally distributed IoT/Edge networks to mission-critical financial platforms, to MLOps pipelines for HIPAA-compliant life sciences.
Proponent of an AI-native development culture, applying a multi-LLM methodology and developing in-house micro-LLMs to architect, document, build, and test the next generation of cloud infrastructure.
Kubernetes MLOps
A modular, production-grade MLOps platform for edge and hybrid Kubernetes clusters built on NVIDIA Jetson Orin and Raspberry Pi devices. It provides GPU-accelerated AI infrastructure with secure, automated setup scripts and clear, reproducible workflows. Designed in a tutorial style, it guides users step-by-step through building a Kubernetes cluster from scratch while offering flexible, customizable scripts to adapt for different hardware, workloads, and environments.
https://github.com/christimahu/k8s-mlops
Words vs Tokens
An experimental visualization exploring the relationship between human language and machine learning tokens. This interactive 3D graph demonstrates the concept of how a sentence, understood by humans through the association of words, can be deconstructed by an AI into a relational map of tokens. It's a conceptual look into how machines "see" language, finding patterns in token relationships in a way that is analogous to how humans find meaning in words.
https://github.com/christimahu/words-vs-tokens
DevRS: Containerized Development Workflow
DevRS is an AI-assisted rewrite of Christi’s long-evolving open source dev environment, supporting extensible development across Rust, Go, Python, C++, and systems-focused languages, as well as modular support for cloud SDKs. Especially useful in consulting contexts, it isolates tooling per project, avoiding version conflicts across diverse client stacks. At its core is Ubuntu on Docker. Before the rise of containers, earlier versions ran on VMs like Vagrant and VirtualBox — even on a Raspberry Pi Kubernetes cluster. Orchestration is implemented 100% in Rust, emphasizing reproducible, Linux-native workflows. DevRS includes preinstalled toolchains, Google Cloud SDK, shell utilities, and a NeoVim setup optimized for cloud development and systems programming.
Blahaj PI: Transgender Sentiment Analysis
Christi developed Blahaj PI, a C++23 machine learning command line interface (CLI) to identify and filter harmful content targeting transgender and non-binary individuals. The project leverages Neural Networks for Natural Language Processing (NLP) sentiment analysis to enable content classification and visualization tools, including a dynamic word cloud, for deeper insights into online discourse. With a focus on performance, accuracy, expandability, and collaborative development, Blahaj PI features configurable models, batch processing capabilities, and detailed confidence scoring for content evaluations. The project is actively evolving, with plans to expand real-time social media scanning and improve accessibility for broader community use.
https://blahajpi.com/
Gai Keep: LLM Content Archiver
Gai Keep (Generated Artificial Intelligence Keeper) is a ground-up rewrite for capturing and organizing responses from Large Language Models (LLMs). It allows users to preserve specific AI-generated content by saving entries as timestamped JSON files, each optionally tagged with user-defined labels.Additional functionality in prototype: behind the scenes, Gai Keep continuously scans the local entry store and applies unsupervised machine learning using neural networks and NLP. These techniques generate new, AI-inferred tags by clustering semantically similar content and surfacing hidden themes or relationships across entries. The system then adds these tags alongside the user’s, helping users reflect on and explore their stored LLM knowledge in novel ways.Though currently in its early scaffolding phase, Gai Keep lays the foundation for a full-featured system that merges AI with personal memory.
https://gaikeep.com/
University of Washington: Machine Learning
Christi expanded her expertise in high-performance computing and artificial intelligence (AI) at the University of Washington. She completed a Machine Learning Certificate program, building and training deep learning models in PyTorch with GloVe embeddings for sentiment analysis, and developing a Generative AI (GenAI) Retrieval-Augmented Generation (RAG) prototype using LangChain, Chroma vector store, and Azure OpenAI. She deployed ML models on Azure infrastructure, managing GPU-enabled Azure VMs with Tesla V100, and explored Hugging Face Transformers for language modeling and fine-tuning. Previously, Christi completed the UW Embedded & Real-Time Systems Programming Certificate, where she gained expertise in low-level systems programming in C, real-time constraints, embedded architecture, and hardware-software integration.
Early Coding: From Chatbots to Networking Exploits
Christi started programming at 11 on a TRS-80 Pocket Computer handed down by her mom. She wrote her first chatbot in BASIC to help pass history tests. By 13 she was building turn-based C++ games; soon after, that curiosity crossed into rule-bending... she wrote a script that spread across her school's network, capturing logins before finally sending a colorful ASCII message to a crush: “Will you go out with me?” The message was well received, but ironically, Christi’s social programming hadn’t advanced as far as her technical skills, and she was too frozen to respond. A year later, she and a friend stumbled onto an unprotected password file at a local ISP and used it to get free internet, learning firsthand how fragile and fascinating connected systems could be. Those early experiments lit the path that would eventually lead to a career spent exploring how machines learn, communicate, and misbehave — and how introducing a bit of that mischief back into the system can build resilience.