kevin@arcelabs — boot sequence
superposition active — click anywhere, or use the toggle above

I build cloud backends that have to stay up — and small sharp tools for the things that annoy me on the way there.

I'm Kevin Arce, AI Native Engineer at TELUS Digital, on the SmartHome+ IoT platform. M.Sc. Quantum Computing in progress at Universidad de La Rioja, creator of Quantum Meowchanics, with an astrophysics side quest at ExoNotes. Scroll to explore the dashboard.

Academic Core

Universidad Francisco Gavidia — by way of Mathematics and Physics

A foundation built on mathematics, physics, and algorithms — the substrate every later backend and cloud decision stands on.

That foundation is what made the jump toward quantum systems a continuation, not a detour.

Universidad de La Rioja, Spain — in progress

Extending a CS + Math/Physics foundation into quantum systems: qubits, superposition, and measurement as the next layer of the same stack. Which means my code now works and fails simultaneously.

Written back out as Quantum Meowchanics .

same substrate, new state space.

Experience Architecture

TELUS Digital AI Native Engineer · CurrentWeb Informática Backend EngineerEnglish4Callcenters Backend EngineerUFG Research Ad-honorem ResearcherFreelance Mobile Independent Developer

TELUS Digital

AI Native Engineer — Current Role

  • Backend, cloud and LLM work on SmartHome+, the flagship IoT platform.
  • Infrastructure as code on AWS with CDK, Lambda and DynamoDB.
  • TypeScript and NestJS services powering connected-home features.
  • Regression and progression test automation protecting production.
  • On-call rotation that occasionally disagrees with my sleep schedule.
  • Agent-first day to day: user stories and a technical spec exist before anything gets delegated.

Web Informática

Backend Engineer

  • Built services in Node.js, TypeScript, and Python.
  • Implemented MySQL Full Text Search for fast in-app querying.
  • Optimized serverless architecture to reduce infrastructure cost.

English4Callcenters

Backend Engineer

  • Built high-throughput services with Bun.js and the Elysia framework.
  • Modeled data in MongoDB and monitored production with Sentry.
  • Integrated Llama 2 inference via Cloudflare Workers AI.

UFG Research

Ad-honorem Researcher — Centro de Modelaje Matemático, Universidad Francisco Gavidia

  • Research and mentorship at the university's mathematical modelling centre.
  • Reinforced the math/physics foundation that later fed into backend and quantum work.

Freelance Mobile Dev

Independent Developer

  • Designed and shipped Android apps independently, published on the Play Store.
  • GermanFromZeroToHero — a Flutter German-learning app, and one of the first things I ever sold.
  • Owned the full cycle: product, UI, backend, and release.

What I Actually Build

Not a skills dump — this is the map of where each thing gets used.

  • Cloud backends

    Build and operate the services behind a production IoT platform. Infrastructure as code, test automation for regression and progression, and an on-call rotation that occasionally disagrees with my sleep schedule.

    • TypeScript
    • AWS CDK
    • Lambda
    • DynamoDB
    • NestJS
  • LLM plumbing

    Wiring models into products real people actually use — moderation bots, assistants, typed workflow graphs. And working agent-first day to day: user stories and a technical spec exist before anything gets delegated.

    • Gemini
    • OpenAI
    • Llama 2
    • Workers AI
    • Copilot
    • Jules
  • Developer tooling

    Small, opinionated CLIs and bots that each delete exactly one recurring annoyance — then actually get published instead of rotting in a folder.

    • TypeScript
    • Rust
    • ts-morph
    • Probot
  • Python & ML

    scikit-learn, dataset building, and the occasional cryptography rabbit hole. Yes, I implemented Diffie–Hellman by hand. No, nobody asked me to.

    • Python
    • scikit-learn
    • Jupyter
    • FastAPI
  • Quantum

    Creator and maintainer of Quantum Meowchanics, first principles to AWS Braket. Where the Maths and Physics background stops being small talk and starts being homework.

    • Braket SDK
    • Qubits
    • Linear algebra
The full toolbox — for the curious and the keyword-searching
Languages
TypeScript · JavaScript · Python · Rust · Dart · Shell
Runtimes & frameworks
Node.js (Express, NestJS) · Bun (Elysia) · FastAPI · Flask · Flutter
Cloud & infra
AWS (CDK, Lambda, EC2, S3, SES, CloudWatch) · Cloudflare Workers · Docker · Kubernetes · GitHub Actions · Turborepo
Data
PostgreSQL · MySQL · MariaDB · MongoDB · Redis · DynamoDB · Firebase
Observability & testing
Sentry · Prometheus · PagerDuty · Jest · Vitest
Ways of working
Agile / Scrum / Kanban · trunk-adjacent PR flow · specs before agents

Currently in Superposition

Two long-running things that aren't side projects so much as a second and third stack.

Quantum Meowchanics

Creator & maintainer

Quantum computing from first principles to AWS Braket — where the cat is in a superposition of understands quantum computing and does not, until you measure it with the exercises.

An offline-first e-learning platform and full curriculum: eight terms that start at classical mechanics and end at running circuits on AWS Braket. Built as a zoneless, prerendered Angular PWA with KaTeX for the maths, Python for the worked examples, and Leitner spaced-repetition flashcards so the material actually sticks.

Terms
8
Lessons planned
127
Hours of material
~654
lessons written 61 / 127

Curriculum spine

  • Done: Pre Classical → quantum mechanics 18 lessons
  • Done: T0 Mathematical & computational foundations 16 lessons
  • Done: T1 Quantum mechanics 19 lessons
  • Pending: T2–5 Core, algorithms, noise & error correction, Braket planned
  • Done: Ref Supplementary textbooks 8 lessons
  • Angular PWA
  • KaTeX
  • Python
  • AWS Braket
  • Offline-first

ExoNotes

Pre-registered study · astrophysics side quest

Do the free-text comments astronomers write on TESS Objects of Interest carry disposition-relevant information that the numeric catalogue columns do not?

A language model turns ExoFOP comments into numeric semantic features; gradient boosting is the predictor; grouped cross-validation by host star is the arbiter. The model's own probabilities are never thresholded into a decision — astrophysics is maximally out of distribution for it. A null result is a valid outcome, and the repo is built to report one honestly.

Labelled TOIs
2,721
Host stars
2,573
Spend to date
~$0.0002

Pre-registered gates

  • Done: G1 Pipeline sanity 0.9154 AUC vs 0.5000 prior
  • Pending: G2 Noise floor not yet run
  • Pending: G4 Semantic lift not yet run
  • Pending: G5 Leakage control not yet run
  • Pending: G6 Public release not yet run

47.9% of the comment corpus restates the label

A planet has a name only because it was confirmed, so a comment reading “WASP-68 b” is the label written in prose. TF-IDF on that text scores 0.969 AUC — that number is leakage, not signal, and must never be cited as evidence for the hypothesis.

  • Python
  • TESS / ExoFOP
  • Gradient boosting
  • Pre-registered
  • ArceLabs My own platform sandbox: NestJS services folded into a Turborepo/Bun monorepo, Kubernetes underneath, Prometheus and PagerDuty on top. Handing myself the on-call problems I don't get to redesign at work.
  • Griffiths, one chapter at a time Working through David J. Griffiths' Introduction to Quantum Mechanics for the MSc — the part where the Maths and Physics background stops being small talk and starts being homework.
  • Agent workflow graphs Modelling LLM steps as typed nodes in a graph instead of one heroic mega-prompt. Specs and user stories exist before anything gets delegated.

Projects

BreakPoint

A GitHub App that fails your PR when you quietly break your own API. It generates OpenAPI schemas for the PR and the base branch, diffs them, and checks that your version bump matches the size of the damage. Split into two layers on purpose: Probot owns GitHub, Actions owns compute.

  • TypeScript
  • Probot
  • OpenAPI
  • GitHub Actions

Botditor

A Reddit moderation bot that's actually live in Reddit's app directory. It ingests comments through the CommentSubmit trigger and runs them past an LLM — because, as its own tagline asks, was that comment written by a human or a slightly sentient toaster?

  • TypeScript
  • Devvit
  • Gemini
  • OpenAI

Envinator

"Come with me if you want to deploy." npx envinator-cli walks your AST with ts-morph, finds every process.env you forgot to document, and interrogates you until your .env is complete. Real parsing, not regex — it won't be fooled by a comment. Masks secrets in its own logs.

  • TypeScript
  • Node.js
  • ts-morph
  • CLI

bun-cli

A Bun project scaffolder written in Rust — my most-starred repo, later torn down and rebuilt around a Result<T, E> core with a modular, cross-platform layout. Shipped to crates.io, where it's been pulled 5k+ times.

  • Rust
  • Bun
  • CLI
  • crates.io

LESSA dataset

About five years ago I built El Salvador's first Salvadoran Sign Language (LESSA) image dataset, because one didn't exist. It's since become one of the most downloaded open datasets to come out of the country.

  • Dataset
  • Kaggle
  • Computer Vision
  • Accessibility

ghostforge

My entire macOS terminal, reduced to one install.sh. Ghostty + Starship + zoxide/fzf/eza/bat/yazi/lazygit, with Catppuccin that follows the system between light and true-black OLED dark. Backs up your .zshrc before it touches a thing.

  • Shell
  • Zsh
  • Ghostty
  • Catppuccin

Linux-Arctis-Manager

Chased a headset pairing bug down to the USB layer of an open-source SteelSeries driver and fixed it upstream: device reset on initialization, re-claiming interfaces after kernel detach. Merged.

  • Python
  • Linux
  • USB
  • Upstream PR

GermanFromZeroToHero

A Flutter German-learning app from the Firebase era, and one of the first things I ever sold. Still the reason I know exactly what a StreamBuilder does.

  • Dart
  • Flutter
  • Firebase
hidden channel — whoami --extended