What I Build

Service categories I can ship to production.

Each line below is a category I have already built and run in production at least once. Where a Stitch + Hyperframes demo exists, it sits inline — designed in the brand of that category, not in the portfolio’s tone.

01

Landing pages with chatbot routing

A landing page wired into a LINE or web chatbot that qualifies the visitor and hands them an AI agent — leading to booking, consultation, or order without a human in the loop until necessary.

Typical shape
LP → LINE → AI agent → booking / consultation handoff
Stack
Next.jsLINE Messaging APILLM

Demo

BotFlow

Convert visitors into bookings

Qualify leads 24/7. Let AI close.

94%
Response
3x
Conversions
24/7
Active
Powered by LINE
🤖
Support Chat
Online
🤖

Hi! Ready to book? I can find you a slot in seconds. 👋

I'd like to book

🤖

Great! Which day works for you?

Mon 10:00
Tue 14:00
Wed 11:00

Mon 10:00

🤖

✓ Confirmed! Mon 10:00 — see you then.

Type a message...

02

Custom internal SaaS

An internal-use SaaS tailored to one specific operational pain — CRM, booking, dashboard, knowledge base. Built around the existing flow, not against it.

Typical shape
Internal CRM, booking app, ops dashboard, knowledge wiki
Stack
Next.jsSupabasePlaywright e2e

Demo

Operate
KH
Customers
Bookings3
Tasks
Reports
Customer
Tanaka Corporation
Enterprise · Tokyo
Last contact
2 days ago
Open deals
¥4.2M
AI Summary
High-value account. 3 open deals ¥4.2M. Last interaction positive — ready for upsell.
Renewal due in 18 days. Sentiment: warm. Recommend scheduling a check-in this week.
Account at risk — no contact in 12 days. Decision-maker on leave. Flag for follow-up.
Notes
#renewal#enterprise#warm
12
Total Customers
3
Bookings Today
87%
Close Rate
Today's Tasks
Send follow-up to Tanaka Corp
Review Q3 proposal draft
Schedule onboarding call
Update deal pipeline
Send invoice #0047

03

AI inquiry concierge across LINE / email / web

A persistent CS layer that classifies intent across channels, drafts contextual replies, escalates the rest, and books calendar slots when appropriate.

Typical shape
Intent routing, multi-language responses, low-rating escalation
Stack
n8nLLMLINE / Gmail / Calendar API

Demo

Inquiry Concierge
AI-unified inbox
LIVE
All
LINE
Email
Web
Open:6
Web
Liu Weijust now

I would like to schedule a demo call...

BOOKING
Reply drafted
LINE
Tanaka M.2m ago

I want to book a consultation for next week...

BOOKING
Reply drafted
Gmail
sarah@acme.io5m ago

Hi, quick question about your pricing plans...

FAQ
Reply drafted
Web
Anonymous8m ago

This service is completely broken and I need...

COMPLAINT
ESCALATE TO HUMAN
LINE
Kimura Y.12m ago

Please cancel my subscription immediately...

SENSITIVE → ESCALATE
ESCALATE TO HUMAN
Classification accuracy 98.4%
LINE
Email
Web

04

SNS automation with brand-aware generation

A SNS pipeline that takes one input (photo, theme, raw note) and produces platform-optimized posts, holds them in a queue for review, and tracks engagement back into the next round.

Typical shape
Multi-platform post generation, scheduling, engagement analysis loop
Stack
n8nClaude Code / OpenClawPostizlocal DB

Demo

BrandPost
Source:
Spring espresso menu launch
Summer limited edition tote bag
Behind-the-scenes roasting day
Instagram
Spring Sip Season is here.

Our new espresso blend wakes up slowly — just like the best mornings should.

Limited summer in a bag.

Our summer tote is here. Lightweight, over-dyed linen. Made for warm days.

From green to gold.

Our roastery, 6AM. The ritual that starts every perfect cup.

#CoffeeAesthetic#SpringMenu#EspressoLover
X / Twitter
New spring menu just dropped ☕ Three origin blends, one limited-edition latte, and a cold brew you'll crave until October. Worth the queue. (Thread 🧵)
We made 50 totes. They're gone by Friday. Oversized, undyed linen, heat-stamped logo. Drop link in bio 👜
6AM in the roastery. Quietly one of the best hours of the week.
#CoffeeTwitter#SpringMenu142 / 280
in
LinkedIn
We revamped our spring menu around one idea: slow mornings.

Sourcing, blend ratios, and the decision to drop two bestsellers. A short write-up on intentional product design in a saturated market.

50 units. Handmade. Sold out in 3 days.

Lessons from our first physical product launch — what we got wrong, and the one thing that worked.

Consistent quality at scale starts at 6AM.

How we redesigned our roasting process to maintain flavor profiles across 4× volume growth.

#ProductDesign#HospitalityIndustry#FoodAndBeverage

05

n8n × LLM workflow builds

Production n8n self-hosted workflows for invoice collection, lead enrichment, competitor monitoring, monthly report generation, OCR — wired to OpenAI / Anthropic / Vision APIs.

Typical shape
Invoice collection, lead enrichment, competitive intel, monthly reporting
Stack
n8nGPT-4o VisionGoogle Sheets

Demo

n8n/ Invoice Collection
Active
Webhook
Trigger
👁
AI Vision
(PDF)
🧠
GPT-4o
Extract
📊
Google
Sheets
Gmail
Reminder
2,847runs / month
99.2%success rate
0.8savg. time
[12:34:18] Webhook received — invoice_batch_0047
[12:34:19] Vision API ✓ 25 fields extracted (3 pages)
[12:34:21] GPT-4o ✓ Structured JSON output validated
[12:34:22] Sheets ✓ Row appended → Sheet "Invoices Q2"
[12:34:23] Gmail ✓ Reminder sent → accounts@client.jp
[12:34:18] Webhook received — invoice_batch_0048
[12:34:19] Vision API ✓ 31 fields extracted (5 pages)
[12:34:18] Webhook received — invoice_batch_0047
[12:34:19] Vision API ✓ 25 fields extracted (3 pages)
[12:34:21] GPT-4o ✓ Structured JSON output validated
[12:34:22] Sheets ✓ Row appended → Sheet "Invoices Q2"
[12:34:23] Gmail ✓ Reminder sent → accounts@client.jp
[12:34:18] Webhook received — invoice_batch_0048
[12:34:19] Vision API ✓ 31 fields extracted (5 pages)

06

Operator + customer two-layer SaaS

A two-interface SaaS where the operator agency runs the approval queue and the customer SPA surfaces results plus next-step input — the same model used in the multi-store MEO build.

Typical shape
Multi-tenant ops console, client-facing SPA, RLS isolation
Stack
Next.jsSupabase RLSPlaywright (admin / client / workflow)

Demo

My Results
TK
Last month
Visibility
Clicks
Reviews
This month setup
Upload this month's photosJPG / PNG, up to 20 files
storefront-april.jpg
Uploaded
Next month live
Approval Queue
1 pending
12
Total
11
Done
1
Pending
Tanaka Shop
Submitted: Just now
NEW
storefront-april.jpg
1 photo · ready for review
Operator

07

Data extraction and shaping

Web scraping, PDF extraction, OCR, API ingest — landing into a database with a clean schema. The point is not to scrape; the point is the shape on the other side.

Typical shape
PDF invoice → bookkeeping sheet, web → DB sync, OCR → structured DB
Stack
PythonPlaywrightAI Vision

Demo

DataShape/ Extraction Pipeline
LIVE
4 rows ingested
SOURCES
PDF
invoice_q1.pdf
214 KB
PDF
receipt_mar.pdf
88 KB
WEB
shopify_api
endpoint
OCR
scan_0042.png
1.1 MB
3 queued
PIPELINE
OCR
Schema match
Validate
throughput
2.4rows/s
OUTPUT
clientamountdatetax_class
Tanaka Corp$8402024-03reduced
Yamada LLC$2,1052024-03standard
Fujita & Sons$??-------
anomaly flagged
Ikeda Trading$5532024-03standard
Nakamura Co.$1,2802024-02reduced
4rows ingested
1excluded

08

Solo prototypes and proofs of concept

Idea validation, MVP design, end-to-end prototypes built in 1–2 weeks. The deliverable is enough running code to make the business question answerable.

Typical shape
Idea validation, prototype, MVP design
Stack
Next.jsn8nPython

Demo

PoC Studio1–2 week builds
ideascopeship
ideas ✦
💬
AI reply bot
validate
📊
MRR tracker
scope
🗺️
Route planner
spike
wireframe ✏
HeaderFeedCardCTA
sketching...
live build 🚀
zsh
npm run dev
▶ ready on :3000
compiled in 1.2s
localhost:3000
running
Loading...
Dashboard v1

Bring the workflow you would describe to a new hire.

Thirty minutes is enough to identify where automation pays back, and to decide together what to build first.