david kalina

Full-Stack Software Engineer · Colorado / Remote

I build scalable
systems and solve real-world
problems—end to end.

I design and ship scalable, high-performing, non-blocking systems — quickly, and with care for the people that use them.

VIEW SELECTED WORKGET IN TOUCHRESUME ↓
Professional exp.
4+yrs
Pipeline cost cut
98%
Saved annually
$20k+
Domains shipped
3+

Recent
work.

Compact case studies on what I built, how I built it, and the impact it had.

01˝

AI Document Processing Pipeline

AI-powered pipeline for TMS software saving $2k+ a month. Extracts structured data from shipping documents in a non-blocking, asynchronous workflow.

1/50th
Token cost
$20k+
Saved annually
100s
Uses per day
vueexpresssqsaws-lambdavercel-ai-sdkterraform
Links
app.cudacartagetms.com
Vue
CLIENT
PDF upload form, job status polling, and the auto-fill flow that saves operators hours of manual entry.
Express
API
Plain REST — endpoints for upload, poll, and hydrate. Form fills the moment the worker finishes writing.
SQS
QUEUE
Upload writes a job row, drops a message, returns in milliseconds. Heavy work happens off the request path.
AWS Lambda
WORKER
Consumes the queue. Rasterizes the PDF with Ghostscript, calls the AI SDK, writes results back to the DB.
Vercel AI SDK
EXTRACTION
Structured outputs turn the page image into typed shipment JSON. No regex, no fragile templates.
Terraform
INFRA
Whole stack as code. Remote state in S3, apply gated behind a GitHub Actions runner — no one ships from a laptop.
USER JOURNEY · Upload → auto-filled form
01
Upload PDFOperator drops a shipping document into the web app.Vue
02
Job queuedRecord written, SQS message fired, response returns instantly.SQS · Express
03
Worker picks upLambda pulls the message and rasterizes the PDF via Ghostscript.AWS Lambda
04
AI extracts fieldsPage image in, structured shipment JSON out.Vercel AI SDK
05
Form auto-fillsFrontend poll resolves; extracted data populates the operator's form.Vue · Express
SHIPPED · 1/50TH TOKEN COST
role: full-stack · pipeline + infra + api
02

Route King

Discovery + tracking app for electric longboarders. Find trails, record rides offline-first, broadcast live to your crew, and compete for segment 'Kings' — best gate-to-gate time wins.

SSE push
Live stream
offline-first
Ride capture
spatial
Every query
react-nativetypescripttrpcpostgisredismaplibre
React Native
CLIENT
Expo Router app — trail discovery, the live ride HUD (elapsed, splits, vibration ribbon), and ride capture that buffers fully on-device so a session is never lost to a dead zone.
TypeScript
LANGUAGE
End to end across the pnpm monorepo — app, Fastify backend, shared Zod schemas. Types flow from the DB to the screen.
tRPC
API
tRPC v11 on Fastify. Location pings are mutations; viewers consume a subscription over SSE — typed end to end, auto-reconnecting, with a react-query polling fallback.
PostGIS
DATA
Rides stored as PostGIS geography with a GIST index. Sync runs authoritative gate-timing on the canonical track and awards segment Kings — every core query is spatial.
SSE
LIVE
Broadcast on by default. Each ping is persisted (late joiners replay) and published to an event bus; spectators render the same server events as the rider and can fire boosts.
Redis
SCALE
Pub/sub seam behind the EventBus interface — swap the in-memory bus for Redis to fan the live stream out across multiple backend replicas.
USER JOURNEY · record → sync → King
01
Start a ridePick a trail and a board. Broadcast goes live to your crew by default.React Native · tRPC
02
Captured on-deviceGPS + accelerometer buffer locally and auto-pause when you stop — nothing lost to bad signal.React Native
03
Crew watches livePings stream over SSE; spectators see the dot move and send boosts.tRPC · SSE
04
Sync reconcilesServer runs gate-timing on the canonical track and awards segment Kings.PostGIS
05
Get your gradePost-ride King Grade: speed-vs-King per segment plus how cleanly you rode.TypeScript · PostGIS
BUILDING · TESTFLIGHT BETA 2026
role: solo · design + eng + realtime

Full-stack engineer building thoughtful systems across SaaS, logistics, and marketplaces.

Four-plus years of professional experience, with a focus on TypeScript, React, and Node.js — and a growing pull toward cloud infrastructure (AWS, Terraform) and AI-powered tooling. Recently, I've taken an interest in cutting down on unnecessary frameworks, packages, libraries, and abstractions while shipping lean, performant software that's easy to reason about.

  • TYPESCRIPT
  • REACT
  • NEXT.JS
  • VUE
  • NODE
  • BUN
  • GRAPHQL
  • POSTGRES
  • REDIS
  • AWS
  • TERRAFORM
  • DOCKER
  • TAILWIND
  • REACT NATIVE
Core skills

TypeScript, React, Vue Node.js, GraphQL, Postgres

Cloud & AI

AWS, Terraform, Docker Vercel AI SDK, MCP

Based in

Colorado Working Hybrid (MT)

Currently

Open to senior roles & select freelance

Featured — 2025

“Designed an AI-powered document pipeline that reduced shipping-PDF processing costs by 98% — saving $20k+ annually.”

EEL DATA SYSTEMSSEE CASE →
Current focus
  • 01AI-powered tooling that amplifies — rather than replaces — human judgment.
  • 02Developing 'anti-doomscroll' mobile software to help people become well-adjusted and more intentional
  • 03Event-driven, non-blocking infrastructure on AWS + Terraform.
  • 04Cross-platform mobile experiences with React Native.

Let's build
something
worth using.

Email — preferred
davidkalina@proton.me
WRITE →
GitHub
github.com/DavidKalina
VISIT →
Location
Colorado · UTC−7
Or — send a note

Prefer email? davidkalina@proton.me