Hello, I'm Macklin.
I'm a software engineer. My lifelong goal is to create experiences that leave a lasting impact.
Projects
NotolibraAn operating system for your notes. No context switching: Annotate PDFs, Ebooks, and save web pages all inside the same app. Integrated AI can operate everything. Your notes are a graph you can navigate and search seamlessly.Rustgpui
- RemlangMobile · JapaneseAn offline-first app with pitch-accent review, kanji breakdowns, stroke-order practice and spaced-repetition decks.React NativeExpoSQLiteWeb · ChineseDictionary search, HSK and community decks, spaced-repetition review and saved readings.Next.jstRPCPostgreSQL
- OdincraftOriginally named Zigcraft, this project has been a longtime vision of mine to create a 3D colony sim with a realistically generated world and endless view distance. Lately, it has become my ongoing benchmark for new LLM releases.OdinSDL3Metal
- Hunter Reeve PoetryA poem website I created for my grandmother, features a poem carousel and podcast player.SvelteKitDrizzlePostgreSQL
Experience
Founding Full-Stack EngineerSacred Texts
Built the content migration, reader and analytics for the September 2026 relaunch of a 20-year-old archive of 3,000+ books, which reached about 295k engaged readers in its first month.
- Migrated 3,071 books (141,765 pages) from hand-written HTML into Postgres, using LLM-first extraction with schema-validated output and per-page provenance.
- Designed and built the book reader, library and category pages, with full-chapter SSR edge-cached on Cloudflare Workers.
- Built first-party analytics and tracing (ClickHouse, OpenTelemetry, Grafana) that separated real readers from bot traffic and found the root causes of launch-week incidents.
SvelteKitPostgresCloudflare WorkersClickHouseOpenTelemetry
Undergraduate ResearcherAIEA Lab, UC Santa Cruz
Studied how large language models reason, reading and presenting more than 30 foundational papers on chain-of-thought and LLM reasoning.
- Read and presented papers weekly, from chain-of-thought prompting and self-consistency to Tree of Thoughts and the Physics of Language Models series.
- Reproduced published GSM8K chain-of-thought results for Llama 3 and Qwen 2.5 within about 2 points, on GPU nodes in the Nautilus Kubernetes cluster.
- Wrote the Kubernetes deployment, volume and secret YAML for those runs, plus scripts to sync code to pods, forward ports and download results.
- Was given access to the Nautilus cluster to continue my own Chinese-to-English fine-tuning project, where Yi-1.5-34B trained on self-curated data reached BLEU 21.73 and chrF 48.00.
PythonPyTorchHugging FaceKubernetesDocker
Writing
Building analytics from scratch
How I built our own observability and product analytics after a website relaunch, using Effect, OpenTelemetry, Grafana and ClickHouse.
How I used LLMs to modernize over 2,000 HTML books
Migrating an internet archive of hardcoded HTML books using structured output, crawling, and a regression-tested feedback loop.
