Skip to content
View gracee3's full-sized avatar

Block or report gracee3

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
gracee3/README.md

Hi, I'm Emmy

I'm a systems and software engineer and solutions architect with 15+ years of professional experience spanning software development, infrastructure, observability, data platforms, compliance, and large-scale technical operations.

I spent seven years at Virtustream / Dell Technologies, ultimately serving as the technical product owner and principal architect for global monitoring and compliance. Today I'm applying that systems background to local AI inference, automatic speech recognition, Rust, hardware-aware optimization, observability, reproducible evaluation, and local-first software.

I've worked with Linux and open-source systems since 2004 and tend to build at the boundaries between software, operating systems, infrastructure, hardware, and measurement.

Resume · Technical portfolio · GitHub repositories

Selected work

Project Focus
native-asr CPU-only offline and streaming ASR with reproducible WER, RTF, latency, and real-time evaluation
qwen38-int8-lab Architecture-aware Qwen3.8-27B W8A8 quantization and validation on dual RTX 3090 GPUs
whisperX-batch Docker-first GPU transcription, benchmarking, and reproducible LibriSpeech evaluation
gpt-oss-rs Rust-native LLM inference, CPU kernels, heterogeneous execution, and evidence-driven performance research
supermicro-observability Secure high-frequency Linux and NVIDIA GPU observability with Prometheus, Grafana, Rust, and NVML
digital-liquid-light-lab GPU-first Rust/wgpu instrument for real-time, physically inspired liquid-light simulation and performance
Mirabile Local-first Rust/Leptos CSR-WASM application architecture
Magnolia Modular real-time DSP, signal processing, plugins, and low-latency systems research

What I work on

I like engineering problems where correctness and evidence matter as much as getting something to run.

That usually means:

  • Rust, Python, Bash, Linux, containers, and systems integration
  • LLM and ASR inference, quantization, benchmarking, and model evaluation
  • latency, throughput, accuracy, memory, and resource measurement
  • explicit architectural boundaries and reproducible test harnesses
  • observability, infrastructure automation, and performance analysis
  • privacy-preserving local and offline systems
  • local-first browser applications using Rust, Leptos, WASM, and IndexedDB

Virtustream / Dell Technologies

At Virtustream, I owned architecture, technical priorities, operational standards, and cross-functional requirements for global monitoring and compliance across enterprise, federal, and Epic healthcare environments.

My work included architecture across 15 multi-tenant colocation sites and more than 7,500 VMs, leadership of a roughly 1 PB HBase/Hadoop-to-BigQuery migration, delivery of a FedRAMP High GCP extension through JAB authorization, observability modernization across VMware and infrastructure platforms, and vulnerability-remediation leadership across more than 2,500 assets.

I also built and scaled automation and operational practices with Ansible/AWX, Python, Bash, Prometheus, Grafana, Zabbix, Kubernetes/K3s, ServiceNow, and related infrastructure, while helping grow the global monitoring support organization from 5 to 25 engineers.

Earlier background

Earlier in my career I worked in enterprise mobile/software engineering, e-commerce production systems, Java, data integration, technical support, and systems engineering.

Engineering approach

Across projects, I try to make the evidence boundary explicit:

define the contract → implement the system → measure it → preserve the evidence → state only what the evidence supports.

That includes publishing negative results and failed gates when they are useful, pinning important dependencies and datasets, distinguishing measured results from planned work, and keeping private data local whenever practical.

For detailed architecture, methodology, and measured results, see the technical portfolio.


Washington, DC · emmygraceclark@gmail.com

Pinned Loading

  1. gpt-oss-rs gpt-oss-rs Public

    CPU-first Rust research implementation of OpenAI GPT-OSS with exact MXFP4 kernels, reproducible benchmarks, and archived accelerator research.

    Rust 1

  2. whisperX-batch whisperX-batch Public

    Docker-first GPU batch transcription and benchmarking harness.

    Python

  3. native-asr native-asr Public

    Offline CPU speech recognition for long-form audio and live streaming: deterministic three-model ensemble plus Nemotron-to-Parakeet cascade.

    Python 1

  4. mirabile mirabile Public

    Rust

  5. qwen38-lab qwen38-lab Public

    Python

  6. supermicro-observability supermicro-observability Public

    Loopback-only Prometheus/Grafana monitoring for a dual-RTX-3090 Supermicro workstation

    Python