MVP · Operations Systems · AI Automation

We turn ideas into web and app products people actually use

We clarify what to build first, then connect design, implementation, launch, and operations through one accountable partner.

1,000+ registered app users
10,000+ public download signal
Public-sector and B2B delivery

Product Engineering

Built to keep working after launch

QUERY RESPONSE TIMEMVNO Hub · production DB15s10s5s0s13.0sBEFORE0.4sAFTER32×FASTERPeak-hour bottleneck query, tuned in production

Live-service improvement

For MVNO Hub, we reduced a bottleneck query from 13s to 0.4s and delivered carrier-order and notification flows.

App releases
Formal releases on both iOS and Android stores
Operations improvement
A critical query improved from 13s to 0.4s
Product scope
Web, app, admin, AI, and cloud operations

A good fit when

You need to think beyond a list of features

Your plan does not need to be complete. We define the problem and priority together, then start with the smallest useful release.

You want to validate an idea as an MVP

Define the core journey people will actually use and launch it as a web or app product.

You want to turn manual work into a system

Connect admin tools, data, and external APIs to reduce repeated work and operating errors.

You want AI inside a real workflow

Turn OCR, document summaries, and search assistance into features with validation, cost control, and logs.

Selected Project Work

Explore the scope and result behind app launches, public and B2B systems, and AI product work.

Methodeck project screen
Own productSelf-planned, launched, and operated

Methodeck

A live personal productivity web service that keeps one shared task pool, reinterprets it through six execution methodologies, and turns verified activity into rewards.

Proof

Built and deployed a Turborepo product spanning Next.js, NestJS, PostgreSQL, shared contracts, and methodology definitions

Next.js 15React 19NestJS 11PostgreSQL 16Drizzle ORMts-rest+5
BudongsanBank Agency Manager project screen
In-house projectReal business work inside an employer

BudongsanBank Agency Manager

A cross-platform business app for real estate agency members, released on both iOS and Android and operated for 1,000+ registered users.

Proof

Released the app on both App Store and Google Play

Cross-platform MobileiOSAndroidApp StoreGoogle PlayB2B Operations
MVNO Hub Renewal project screen
Client deliveryContracted and delivered for a paying client

MVNO Hub Renewal

A public-service renewal project that delivered telecom order integration APIs, order-status Kakao AlimTalk notifications, and a 32x Oracle query improvement.

Proof

Delivered real-time order integration APIs for external telecom operators

JavaSpringOracleRESTful APIKakao AlimTalkGit
Status: Delivered
Visit
Read Case Study
JodalN project screen
Own productSelf-planned, launched, and operated

JodalN

A production-grade public procurement SaaS that collects Korean G2B bid data, summarizes attached documents with LLMs, and supports high-speed search.

Proof

Cloud SQL snapshot (2026-08-26): recorded 135,997 instrumented user activities and 830,889 live bid-notice rows

Next.js 15NestJS 11TypeScriptPrismaPostgreSQLLLM+4
Status: Live
Read Case Study

Open-source libraries

Turn proven product foundations into public libraries

Rather than keeping recurring engineering boundaries inside one product, we publish them as TypeScript packages with install paths, platform boundaries, focused examples, and API references.

10 documented packages

Check it yourself

Most numbers on this site can be verified externally

The hardest part of hiring a one-person studio is that every claim is self-reported. Everything below is a public record you can open right now.

  • Public app store metric

    BudongsanBank Agency Manager - 10,000+ downloads on Google Play

    Open Google Play listing
  • Delivered service running today

    MVNO Hub - a public-facing service you can open and use

    Open mvnohub.kr
  • CNCF certification

    Kubestronaut - all five Kubernetes certifications (CKA, CKAD, CKS, KCNA, KCSA)

    About the Kubestronaut program
  • Reappointed three times

    Naver Connect Foundation Boostcamp backend code reviewer - 2022, 2023, 2025

  • Open commit history

    Actual working history and open-source contributions on GitHub

    github.com/apeltop
  • This site's measured performance

    We sell Core Web Vitals, so measure this site first rather than taking our word for it

    Run PageSpeed Insights

How we work

Turn an open-ended request into a workable delivery scope

We share the decisions that matter at each stage to reduce uncertainty around timing and outcomes.

  1. 01

    Clarify the problem and priority

    Identify users, operating context, must-have functionality, expected usage, data shape, and external API cost before agreeing on the first release scope.

  2. 02

    Design and build

    Design and build the screens, data, integrations, and operating requirements as one flow.

  3. 03

    Validate and launch

    Check critical scenarios and prepare deployment, store release, and the operating environment.

  4. 04

    Operate and improve

    Use logs and real usage to identify issues and set the next improvement priority.

Why work with Ptokos

Start with the problem

We clarify user flows, operating constraints, expected usage, and recurring costs instead of simply translating a feature list into screens.

Build as one system

Frontend, backend, database, and deployment are connected as one product system.

Prepare for after launch

Performance, testing, deployment, logs, and handover are prepared for real operation.

What We Build

Services

We define a practical delivery scope around your current problem. Each service details its scope and related work.

Next.js Website and Admin Development

Build search-friendly web products, admin consoles, and dashboards with Next.js App Router.

What this makes possible

Pages become discoverable and indexable

Next.jsReactTypeScriptTailwind CSSVercel
View Detail

React Native App Development

Build iOS and Android products with Expo, React Native, and a production-minded mobile architecture.

What this makes possible

Launch both stores from one codebase

React NativeExpoTypeScriptiOSAndroid
View Detail

AI and LLM Automation Development

Connect document summarization, OCR, structured generation, and search assistance into real product workflows.

What this makes possible

Automate repetitive document processing

LLMOpenAI APIClaude VisionOCRNestJS
View Detail

Cloud and DevOps Setup

Prepare deployment, logs, tests, and infrastructure automation so MVPs can become live services.

What this makes possible

Reduce manual deployment risk

GCPCloud RunDockerTerraformKubernetes
View Detail

Project scale

How big does a project need to be

A firm quote follows a fixed scope. To help you size things before a call, here is how past projects group into three bands.

Smallest useful release (MVP)

Timeline: About 4-8 weeks

Typical scope
One or two core user journeys, a single platform, and a basic admin screen
Real example
Thanks Note - from planning to App Store release in 2 months

When you need real users in front of the idea before investing further

Operations system or app launch

Timeline: About 2-4 months

Typical scope
Two or more of web, app, and admin, external API integration, auth and roles, and store review
Real example
BudongsanBank Agency Manager - released on both stores and operated for 1,000+ registered users

When the system will be used daily by staff or customers

Operating platform or data pipeline

Timeline: About 4 months and up

Typical scope
Multi-service architecture, data and LLM pipelines, performance and cost tuning, operational logging and deployment automation
Real example
JodalN - 240+ field data models, 19 Prisma models, and an LLM document summarization pipeline

When data volume or traffic means the operating architecture comes first

Price follows the agreed scope rather than the band. If you already have a budget, we start by proposing what is achievable inside it, and we say so directly when it is not a fit.

The person you consult is the person who delivers

GiJu Kang

Independent Product Engineer

Founding member of CNCK, the official CNCF community in Korea - Community Ops Sub-leader

I have built and operated public and B2B operations systems, mobile products, and data collection services. I personally carry the context from consultation through design, implementation, and deployment, without a handoff layer in between.

Directly accountable for and how I work

  • Consultation, delivery scope, UI/API/data design, implementation, and deployment as one continuous flow
  • Plain-language updates that explain the reasoning, options, and delivery or operational impact behind each decision
  • Evaluating AI and OCR against real input documents for accuracy, turnaround time, and cost before connecting them to a product workflow
  • Testing, logging, deployment, and maintenance handover criteria for operation after launch
  • Appointed three times as a Naver Connect Foundation Boostcamp backend code reviewer (2022, 2023, 2025)

Cloud and Kubernetes certifications

CNCF

CNCF Kubestronaut

Awarded by the CNCF to engineers holding all five Kubernetes certifications. Earned as the 104th Kubestronaut in Korea, covering security (CKS, KCSA) as well as administration and development.

About the Kubestronaut program
CNCF
CKACKADCKSKCNAKCSA
AWS
Solutions Architect AssociateDeveloper AssociateCloud Practitioner
Google Cloud
Associate Cloud Engineer
Other
TensorFlow Developer CertificateISTQB Foundation Level

Questions before you start

What people ask before starting, and before signing

Before we start

Can we talk before the requirements or design are complete?+

Yes. Share the problem, intended users, and current pain point, and we can define the first release scope together.

Can web, mobile, admin, and AI functionality be handled together?+

Yes. We prioritize the necessary work, then connect the product surface, API, data, and deployment into one delivery flow.

Can you improve an existing product or a performance issue?+

Yes. We first inspect the current structure and reproducible issue, then propose the impact area and improvement order.

What should I prepare for a first conversation?+

Your project goal, target timing, and any reference service or material are enough. Budget and detailed scope can be clarified together.

Before you sign

You are a one-person studio. What happens if you get sick or overloaded?+

This is the most common question, and a fair one. Three things cover it. First, code and infrastructure are set up under your accounts and repositories from day one, so another engineer can pick the work up at any point. Second, design decisions, deployment steps, and environment variables are documented alongside the work. Third, if anything prevents normal progress for more than three business days, we tell you immediately and you choose between rescheduling and terminating with a settlement.

How many projects do you run at once?+

At most two at a time. During the first conversation we tell you what is already in flight and when we can start. If schedules would collide, we propose a realistic start date or decline rather than overcommitting.

How do you communicate when the schedule slips?+

We share a written progress update once a week and flag a delay as soon as it becomes likely, never on the due date itself. If the delay is on our side, we adjust the schedule at no extra cost. If it comes from a requirement change, we propose the revised scope and timeline and agree it with you before continuing.

What if the result is not what I expected?+

Reviewing only at the end is what creates that problem. We ask for a checkpoint each time a core screen or flow is ready, so work does not run long in the wrong direction. Anything built differently from what we agreed is corrected at no cost. If the requirement itself changed, we re-estimate the scope and timeline together.

How long is the warranty period?+

Thirty days of free defect fixes after delivery. Anything that does not behave as agreed is corrected at no cost within that window. New features or ongoing operations are arranged as a separate maintenance agreement, and you receive the handover documentation and deployment procedure either way.

Can we sign an NDA?+

Yes. Ask at any point, including before the first detailed conversation, and we sign before discussing specifics. Material you share is used only for the proposal and the work itself, and nothing is published as a case study without your prior agreement.

Who owns the code and deliverables?+

Ownership transfers to you on final payment. Repositories, deployment environments, and third-party service accounts are either created under your organization or transferred at handover.

Project consultation

Start by confirming the first release your project needs

Tell us what you want to build and where things stand. We reply within one business day with a workable scope and the next step.

Reply within one business day (24 hours)

The consultation is free, and the reply is a summary of workable scope and expected timing with no sales pressure. If a business day passes with no reply, the message did not reach us - please resend to sunshine@ptokos.com.