sunmoon.dev

AI software studio · Sunmoon.co Pte. Ltd.

Your AI product, shipped in six weeks.

I'm Seunghun Lee — ex-Spotify, ex-Klarna, founder of transcribe.so. I design, build, and ship production AI end to end: assistants that cite their sources, search that understands meaning, products that bill from day one. You own every line.

Book a callSee the work
Ex-SpotifyEx-KlarnaEx-Supabase Expert PartnerWorked with Y Combinator–backed

Flagship product

transcribe.so
#4
of 80+ models on the ASR leaderboard
67
languages in production
3
ASR models, auto-routed per file

Built and shipped by an engineer who has done it at scale

  • Spotify
  • Klarna
  • Worked with Y Combinator–backed startup
  • Ex-Supabase Expert Partner
  • transcribe.so
  • goodlisten.co

The real cost of the wrong build

Most AI projects don't fail at the model. They fail at shipping.

The demo works. The deck looks great. Then the prototype meets real users, real data, and real costs — and it never makes it to production. The agency rotates engineers, the freelancer disappears, and your team inherits code nobody owns.

Meanwhile the models keep improving. The risk isn't picking the wrong model — it's spending two quarters not shipping anything your customers can touch.

Per failed prototype

3–6 months with an agency or contractor

+ $50k+ for a demo that never reaches production

+ No evals, no monitoring, no cost controls

+ A codebase your team can't maintain

= Back to square one, two quarters later

Per quarter of delay

Competitors ship AI features your customers ask you for

+ Your team burns cycles evaluating vendors instead of building

= The market moves while the project stalls

What I build

Production AI, not prototypes.

Three ways to work together. Every engagement is hands-on — the person who scopes it is the person who writes the code and ships it.

AI assistants that know your business

Grounded in your data, able to take action, and honest when they don't know. The same engine behind transcribe.so's cited answers.

  • Cited Q&A with real source attribution
  • Tool-use and function calling
  • Multi-model routing (OpenAI, Qwen, Mistral)
  • Eval harness so quality doesn't regress

Full-stack SaaS, end to end

From schema to Stripe to ship. The product gets built by someone who has shipped and runs his own.

  • Next.js + Postgres / Supabase
  • Auth, organizations, and billing
  • B2B and B2C ready
  • Industry-standard stack, no vendor lock-in

Fractional CTO & AI consulting

Architecture reviews, model selection, and hands-on building. A CTO for hire who writes the code.

  • AI architecture & retrieval review
  • Model and cost selection
  • Hands-on implementation
  • 6 months of email support on builds

Selected work

Products I've shipped and run.

Not case studies for clients I can't name — products I designed, built, and operate myself, plus senior engineering at companies you know.

Flagship · Founder & Engineer

transcribe.so

An AI knowledge library for lectures, podcasts, and videos. Multi-model speech-to-text, timestamped chapters, cited Q&A, subtitle export, and an AI-agent API that drops into ChatGPT, Claude, and Cursor. Ranked #4 of 80+ models on the ASR leaderboard. Designed, built, and operated solo.

  • Multi-model ASR auto-routed per file
  • Cited Q&A with timestamped sources
  • 67 languages
  • AI-agent API + MCP server
transcribe.so homepage — the flagship product built and operated by sunmoon.dev
Founder & Engineer

goodlisten.co

AI-powered podcast discovery plus a creator studio that turns long episodes into highlights, chapters, and clips in one click.

Senior Full-Stack Engineer

Spotify

Freemium — internal tooling and growth surfaces for one of the largest audio platforms in the world.

Senior Engineer

Klarna

Tooling Product Integrations — full-stack across the stack, pushing new tools and better ways of working.

Early Engineering

Y Combinator–backed startup

Hands-on product engineering on a venture-backed product during its early scaling.

Inside the flagship build

What shipping looks like: transcribe.so

Real screenshots from the product I designed, built, and operate. Three steps from raw audio to cited answers — the same engineering standard your build gets.

  1. 1

    Drop in audio or a link

    Paste a YouTube URL or upload a file. The pipeline picks the best speech-to-text model for the language and splits hours of audio into durable background jobs.

  2. 2

    Get structure, not a wall of text

    Every transcript comes back with chapters, topic summaries, and timestamps — generated, checked, and assembled by the same kind of eval-tested pipeline I build for clients.

  3. 3

    Ask anything, get cited answers

    Questions come back grounded in the recording, with timestamps you can click to verify. This is what 'an assistant that knows your business' means in practice.

Try it live
Seunghun Lee

A note from the maker

Hey, I'm Seunghun 👋

In 2023 I left Spotify to work on finding the useful 90 seconds inside a three-hour podcast. We built goodlisten.co, ran out of runway, and I went back to a desk job.

But I kept building. transcribe.so came out of the same itch — I wanted production AI that actually shipped, not another prototype that dies in a notebook. Now I design, build, and run it solo: multi-model speech-to-text, cited Q&A, an AI-agent API, 67 languages.

sunmoon.dev is my studio (Sunmoon.co Pte. Ltd.). I take on a small number of builds for teams who need the same thing — production AI shipped properly. You work directly with me, not an account manager and a rotating bench of contractors. If I take your project, I'm the one writing the code.

Seunghun

2

AI products shipped & operated

67

languages in production

10+ yrs

shipping full-stack software

1

engineer on your build — me

How it works

From idea to shipped, without the drama.

01

Scope

A call to understand the problem, the data, and the constraints. You leave with a concrete plan, a timeline, and a fixed-ish quote — not a vague proposal.

02

Build

I build in the open with short feedback loops. You see working software every week, not status decks. Architecture choices favor your team owning it later.

03

Ship

Deployed, monitored, and documented on an industry-standard stack with no vendor lock-in. A basic agent ships in 4–6 weeks; larger builds in 3–6 months.

04

Support

Six months of email support on every build, plus a discounted on-call package if you want me on speed-dial as you scale.

Who it's for

Built for teams that need it shipped, not studied.

Whether you're going zero-to-one on an AI product, adding AI to something that already works, or need senior technical judgment on tap — the same builder handles it.

Founders going zero-to-one

You have the idea and maybe the funding. You need an AI MVP real users can touch in weeks — built on a stack your future team will thank you for, not a prototype you'll throw away.

Product teams adding AI

Your product works; now customers expect AI in it. Get assistants, semantic search, or speech features integrated into your existing codebase with evals and cost controls — without distracting your core team.

Companies needing a fractional CTO

You need senior technical judgment — architecture reviews, model selection, build-vs-buy calls — from someone who ships and runs production AI, not a slide-deck consultant.

References

What engineering leaders say.

Seunghun has done a great job building internal software, provided positive energy and teamwork. He's a great team player and is very knowledgeable as a full-stack engineer. I highly recommend Seunghun.
Leonardo Teixeira MenezesEngineering Manager, Spotify
Seunghun is known for his technical expertise, adaptability, and innovative mind — well-proficient across frontend and backend, moving between them based on the team's needs. His positive attitude and dedication consistently led us toward effective solutions.
Alessandro d'EsteEngineering Manager, Klarna
Seunghun demonstrated a high level of proficiency and commitment, with exceptional skills in teamwork and problem-solving. He's always looking ahead, ready to tackle the next big thing, and reachable in times of need.
Linus LångbergSenior Software Engineer, Klarna
Throughout our collaboration, Seunghun consistently demonstrated a great ability to generate innovative ideas and contribute positively to the project and team. His dedication and creativity make him an asset to any team.
Desireé BjörkmanSenior Engineer, Klarna

Building in public

Follow the build.

Build logs, product clips, and what people say about the work — across X, Instagram, and TikTok.

X@shsunmoon

Build logs, shipping notes, and AI engineering threads.

Instagram@shsunmoon

Behind the scenes of running two AI products solo.

TikTok@shsunmoonlee

Short clips: demos, experiments, founder life.

Engagements

Transparent pricing. No hidden costs.

Starting points, not fixed menus. Final scope and price come out of the first call.

Consulting / Fractional CTO

from $200/ hour

A CTO for hire who writes the code.

  • Architecture & AI strategy review
  • Model and cost selection
  • Hands-on implementation
  • Bug fixes and rescue work
Book a call
Most popular

AI product build

from $10k/ application

A production AI assistant grounded in your data.

  • B2B / B2C SaaS ready, orgs + billing
  • Cited Q&A and source attribution
  • Industry-standard stack, no lock-in
  • 6 months of email support
Book a call

Full-stack SaaS

from $10k/ application

A complete SaaS build, schema to Stripe.

  • Next.js + Postgres / Supabase
  • Auth, organizations, billing
  • Industry-standard stack, no lock-in
  • 6 months of email support
Book a call

Not sure which fits? Book a 15-minute intro and we'll figure it out.

Questions

Frequently asked.

Let's talk

Have something that needs shipping?

Tell me what you're building. If it's a fit, you'll get a plan and a quote — if it isn't, I'll tell you that too, and point you somewhere better.

Book a 15-min introBook paid consulting

[email protected]X / TwitterLinkedIn