Available for contract · Full-time

Full-stackengineer &AI specialist

Available for contract and full-time roles, shipping reliable products from interface to infrastructure.

Selected work

Projects that ship

Product and systems work across full-stack and AI, each broken into the surfaces that matter.

01/Full-stack · AI

Inference platform

End-to-end product surface for model access, API keys, usage metering, and routing across a multi-tenant inference stack.

Expand details

Dithered Janus marble bust

Console · keys

01

Key & tenant console

Operators create scoped keys, set rate limits, and rotate credentials without touching the API layer. Audit trails land in the same view as live usage.

Metering · charts

02

Usage & cost metering

Token and request metering rolls up by model, team, and environment. Dashboards surface spend anomalies before they hit the invoice.

Routing · graph

03

Model routing

Traffic can fall back across providers and model versions with weighted routes, latency budgets, and kill switches for bad deploys.

02/Systems · AI

Agent tooling

Durable workflows and operator interfaces for multi-step agent runs, with retries, traces, and human handoff baked into the runtime.

Expand details

Dithered multi-armed Hercules sculpture

Timeline · trace

01

Run timeline

Every agent step is recorded with inputs, tool calls, and timing. Operators scrub the timeline to find where a run drifted.

Orchestration · flow

02

Durable orchestration

Long-running jobs survive restarts with checkpoints and idempotent tool execution. Failed branches can resume without replaying the whole graph.

Handoff · queue

03

Human handoff

When confidence drops, the run pauses into a review queue with context packed for a human decision, then continues from the same state.

03/Full-stack · Web

Product rebuild

Performance and UX pass on a production customer-facing app: faster shells, clearer flows, and measurable drops in abandon rate.

Expand details

Dithered Vulcan craftsman bust

Perf · waterfall

01

Shell performance

Critical path assets were split and deferred so first paint and interaction landed under target budgets on mid-range devices.

Checkout · flow

02

Checkout clarity

Form friction and error states were rebuilt around real failure modes. Drop-off between cart and confirm fell without changing pricing.

RUM · dashboard

03

Observability loop

Product metrics and RUM sat next to release markers so regressions showed up in the same week they shipped, not a quarter later.

04/In progress · AI

Personal research bench

A local workspace for evaluating models, prompts, and traces, built for quick comparison loops rather than production traffic.

Expand details

Dithered Athena marble bust

Diff · prompts

01

Prompt diff board

Side-by-side prompt variants with shared fixtures so scoring stays comparable across model swaps.

Traces · local

02

Local trace store

Runs stay on disk with searchable metadata, useful when cloud logging is overkill for exploratory work.

05/In progress · Product

Contract ops kit

Lightweight tooling for scoping, handoff, and delivery cadence across independent engagements.

Expand details

Dithered Roman handshake relief

Scope · canvas

01

Scope canvas

Constraints, milestones, and out-of-scope notes live in one place so kickoffs stop resetting every week.

Handoff · pack

02

Handoff pack

A repeatable export of decisions, open questions, and repo pointers for when a engagement pauses or transfers.

About

Building with clarity

Alex Gwak is a full-stack engineer and AI specialist working at the intersection of product design, systems architecture, and applied machine learning. He builds end-to-end software: interfaces people trust, APIs that stay coherent under load, and infrastructure that can be reasoned about months later.

His practice spans greenfield products and production rebuilds. He has shipped inference platforms for model access and usage metering, agent tooling for durable multi-step workflows, and customer-facing applications where performance and clarity of UX were the primary constraints. Across those engagements, the through-line has been the same: reduce ambiguity, ship reliable increments, and leave systems that other engineers can extend without archaeology.

Alex works comfortably from the browser to the database and into the surrounding operational surface: auth, observability, deployment, and the quiet glue that keeps a product honest in the wild. On AI-adjacent work he focuses on applied systems rather than novelty for its own sake: evaluation loops, traceability, guardrails, and product surfaces that make model behavior useful to real users.

He is available for contract engagements and full-time roles, with a preference for teams that care about pace without sacrificing craft. Collaboration tends to work best when the problem is concrete, the constraints are named early, and the goal is software that holds up after the demo.

Outside of client delivery, Alex keeps a small incubation bench of personal tooling: research environments, contract ops experiments, and prototypes that may later become products or simply sharpen the next engagement.

Experience

Path so far

A concise timeline of roles across product, systems, and applied AI.

01 / 04
  1. 2024 to Present

    Independent engineer

    Contract · Remote

    Engaged across inference platforms, agent workflows, and product rebuilds, owning delivery from interface through infrastructure for startups and mid-market teams.

    Engagements
    6
    Throughput gain
    3.2×
    Uptime held
    99.95%

    Projects

    • Inference console

      Keys, usage metering, and model routing UI for a multi-tenant API.

    • Agent runtimes

      Durable multi-step workflows with retries, traces, and human handoff.

    • Checkout rebuild

      Performance and UX pass that cut abandon rate on a live commerce flow.

    • TypeScript
    • Next.js
    • Python
    • Postgres
    • AI SDK
    • Temporal
  2. 2022 to 2024

    Staff product engineer

    Northline Systems

    Led full-stack delivery for a customer-facing platform: redesign of core flows, API consolidation, and reliability work under production load across a team of eight.

    p95 latency
    −38%
    Activation
    +27%
    APIs retired
    12

    Projects

    • Workspace v2

      Rebuilt the primary product shell with shared navigation and faster load paths.

    • Billing fabric

      Usage-based invoicing, entitlements, and audit trails for enterprise accounts.

    • Observability push

      Unified traces, SLOs, and on-call runbooks for the customer-critical path.

    • React
    • Node
    • GraphQL
    • Postgres
    • Redis
    • AWS
  3. 2020 to 2022

    Full-stack engineer

    Cascade Labs

    Built early product surfaces and internal tooling for applied ML teams: eval harnesses, trace views, and the auth/ops layer around experimental models.

    Eval throughput
    Internal tools
    18
    Setup time
    −60%

    Projects

    • Eval bench

      Batch scoring, prompt diffs, and regression gates for model releases.

    • Trace explorer

      Step-level inspection for agent and pipeline runs used by research and CS.

    • Access layer

      SSO, scoped keys, and environment isolation for sandbox vs production.

    • TypeScript
    • Python
    • FastAPI
    • React
    • Kafka
    • GCP
  4. 2018 to 2020

    Software engineer

    Harbor Digital

    Shipped web products end to end: React interfaces, Node services, and the deployment path that kept weekly releases boring for a growing product team.

    Release cadence
    Weekly
    Bug escape
    −45%
    Products launched
    3

    Projects

    • Client portal

      Account management, document exchange, and status tracking for B2B clients.

    • Design system

      Shared component library and tokens that cut new-page build time in half.

    • CI / CD spine

      Preview environments, smoke tests, and one-click promote to production.

    • JavaScript
    • React
    • Node
    • PostgreSQL
    • Docker
    • Heroku

Contact

Let's talk

Send a note, request a capability package PDF, or book a consultation.

Direct message

Send a direct note via email.

Capability package

Request a capability overview PDF.

Request a PDF package with background, focus areas, and engagement details.

Consultation

Book a consultation on the calendar.

August 2026

Mon

Tue

Wed

Thu

Fri

Sat

Sun

Select a day

Choose an available day to see consultation times.

Contract · Full-time · Consulting

Let's build

Available for contract engagements and full-time roles. Full-stack product, systems, and applied AI. Built to ship and hold up after the demo.

Get in touch

Portfolio 2026

Alex Gwak

Available · 2026