Open for freelance & remote work

Ulil Abshor

Automation & Python Developer

Web Scraping & Bot Specialist

I build scrapers, bots and desktop tools that take repetitive work off people's hands — then ship them tested, packaged and documented.

  • Python
  • Web Scraping
  • Bots & Automation
  • Desktop Apps
  • Cloud Storage
Scroll

About Me

divider

I build automation and AI tools that take repetitive work off people's hands.

Hi, I'm Ulil Abshor — a self-taught engineer. I work across automation, AI agents, web apps, data pipelines and software testing: I learn how a system works, find where the time is being wasted, and build the smallest thing that removes it.

🧰 What I've built so far: an AI admin for online shops (WhatsApp chat → orders → Google Sheets), a private-API automation bot running 112 cycles a day without a failed batch, a 40-source data pipeline with zero duplicates, a document chatbot grounded in the user's own files, and Playwright test suites for bug audits. Different problems, same approach.

🛠️ Stack: Python, JavaScript, LLM APIs, Google Sheets, WhatsApp & Telegram bots, Playwright, SQLite, cron, Linux. Every build ships with logging, alerts, a tested end-to-end path, and a handover doc — because a half-working tool is worse than none.

🚀 Open to automation, AI agents, web apps, data pipelines and QA work — freelance or full-time.

My Skills

divider

Frontend Development

HTML, CSS, JavaScript
Responsive Design
WordPress, Elementor
Git & Version Control

AI & Automation

Python
Claude API / OpenAI API
RAG Systems
n8n Workflows

Operations & Administration

Microsoft Office
Administration
Customer Service (English)

Services & Pricing

divider

AI agents that do the admin work. Fixed scope, fixed price, working in days — not a proof of concept.

AI Admin for Online Shops

3–5 days · fixed scope

A shop admin that answers customers 24/7 and writes every order into a spreadsheet — without copying anything by hand.

  • Answers from your own catalog & FAQ (never invents a price)
  • Product catalog with live stock, one item per line
  • Guided order form — one question at a time, never half-filled rows
  • Chat order → clean row in Google Sheets
  • WhatsApp confirmation + H+1 / H+3 follow-ups
  • Owner bot for reports — customers can't see revenue

AI Social Autopilot

4–6 days · fixed scope

Drafts your daily posts from your product data, waits for your approval, then schedules them — and reports what people said.

  • 3–5 drafts per day in natural Indonesian
  • One-tap approval queue in Telegram
  • Scheduled posting via official APIs
  • Daily digest of mentions & comments

AI Operations (bundle)

7 days · admin + content

Admin and content running together, with one daily report and one person responsible for both.

  • Everything in AI Admin + Social Autopilot
  • One combined 07:00 report: orders, revenue, content
  • Monthly care: monitoring, small fixes, backups
  • Priority support on WhatsApp
Discuss your project on WhatsApp

Scope fixed before we start · 50% upfront, 50% on delivery · 2 free revisions

⚡ Limited slots — I take 3 new builds at a time so each one ships on schedule.

Every delivery is tested end to end before handover: customer chat → spreadsheet row → WhatsApp confirmation. If a step fails, it is not finished.

Selected Work

divider

12 shipped builds across automation, scraping and QA. Click any card for the full diagram.

Standalone Desktop Manga Scraper & Cloudflare R2 Pipeline

Category: Desktop App · Automation · Cloud Storage · Web Scraping

Tech: Python, CustomTkinter, PyInstaller, Cloudflare R2 (S3 API), Pillow (WebP), HTTPX, BeautifulSoup4

Windows GUI app (dark mode) that scrapes three manga sources, converts every page to WebP (quality 65) and multipart-uploads it straight to Cloudflare R2.

✅ Non-blocking background threading · zero-dependency portable .exe · hierarchical output with HTML-tagged chapter manifests for WordPress / MangaStream

Desktop manga scraper GUI: sources, HTTPX parser, Pillow WebP q65, Cloudflare R2 multipart upload, local HTML manifest

AI Document Chatbot

Tech: Python, Streamlit, Claude API, FAISS

Upload PDF/TXT/DOCX — ask questions, get answers grounded in the document content.

Demo: Screenshot below

AI Document Chatbot Demo

Auto-Apply Bot

Tech: Python, SMTP, Excel, Gmail

Automated job application bot that reads job listings from Excel, filters vacancies, and sends applications via Gmail.

✅ 12 applications sent out of 33 vacancies

⏱️ Automatic delay between applications

Auto-Apply Bot Demo

n8n WhatsApp AI Chatbot

Tech: n8n, Qdrant, Hugging Face, Google Drive, WhatsApp API

AI-powered WhatsApp chatbot with RAG system — answers user questions based on uploaded documents.

n8n WhatsApp AI Chatbot Workflow

Automation Bot — 100+ cycles, 0 failures

Tech: Python, REST API, wallet signature auth, cron, SQLite

Reverse-engineered a browser game's private API, replaced manual clicking with a scheduled bot: harvest, replant, water, craft, sell, auto-upgrade.

✅ 100+ scheduled cycles in a single day · 0 failed batches

⏱️ Daily digest + failure alerts to Telegram

Automation bot dashboard: 112 cycles in 24h, 0 failed batches, action log

AI Admin — chat, order, Sheets, WhatsApp

Tech: Python, LLM extraction, Sheets API, WA gateway

Shop admin that answers from the shop's own catalog, guides buyers through an order form, and writes each order to Google Sheets.

✅ ~1.5s replies · live stock · duplicate-proof (retries deduped) · owner-only reports

AI Admin flow: customer chat, Google Sheets order row, WhatsApp confirmation

Portfolio Website

Tech: HTML, CSS, JavaScript, Bootstrap

Personal portfolio website — you're looking at it right now!

🔗 Live Demo

Portfolio Website Screenshot

Python Web Scraper with Proxy Rotation

Tech: Python, Web Scraping, Proxy Management, JSON

Resilient Python scraper with automated proxy rotation and fallback logic to prevent rate limits. Built with multi-worker concurrency, 2s auto-timeout failover, and structured JSON export.

Terminal log of the Python scraper: proxy pool init, proxy timeout, direct fallback, parsed payload

n8n AI Workflow Automation Pipeline

Tech: n8n, Claude API, [OI] API, Webhooks, CRM Automation

Production-ready operational AI pipeline built in n8n for automated lead triage and routing. Handles webhook normalization, LLM intent classification, CRM/Google Sheets sync, and dead-letter fallback.

n8n workflow architecture: webhook trigger, validation, LLM router, switch, CRM sync, error fallback

Automated Video & Audio Processing Script

Tech: Python, FFmpeg, Whisper, Audio Processing

Automated multimedia pipeline: voice + music mixing with intelligent ducking, music trimmed to voice length, and Whisper captions burned frame-accurate into the video.

✅ −10.5 dB music ducking · captions per sentence · one-command render

▶ Play demo video

Frame from the generated demo video with burned-in Indonesian captions

Web3 dApp Functional QA & Regression Test Log

Role: Lead QA Tester · Category: QA & Web3

End-to-end regression and transaction-integrity matrix: wallet session recovery, DEX swap slippage handling, deep-link transaction parameters, network-switch guards, RPC/CEX balance sync.

✅ 42 test cases · WalletConnect v2, Rabby, MetaMask, Trust Wallet · Blocker→Minor defect log with expected vs actual

📄 Preview PDF  ·  ⬇ Download

Web3 dApp QA test log: 8 defects across wallet sign-in, DEX swap, signature timeout, deep-link, CEX sync, approvals, network switch

Edge AI Motion Classifier & CI4 Pipeline

Tech: MediaPipe · CodeIgniter 4 · WebAssembly · Computer Vision

Pose-based motion classifier running fully in the browser: 33 3D landmarks, sub-25 ms inference, zero video streaming to the server.

✅ jump · crouch · kick · lying down · 300 ms state lock

Read the case study ↓

Edge AI motion pipeline diagram: webcam, getUserMedia, MediaPipe Pose Landmarker, kinematic classifier, CodeIgniter 4 view

Edge AI Motion Classifier & CodeIgniter 4 Pipeline

divider

Zero-streaming backend · sub-25 ms edge inference · MediaPipe Pose Landmarker (33 landmarks, 3D) running entirely in the browser.

EXECUTIVE SUMMARY

A pose-based motion classifier that runs where the camera is. The browser captures frames with getUserMedia(), runs the MediaPipe Pose Landmarker through WebAssembly with the GPU delegate, and classifies movement from 33 3D landmarks — all before a single byte leaves the machine.

The CodeIgniter 4 layer does one job: render a clean view wrapper. There is no frame proxying, no video queue and no media worker on the server. Only state events (state, confidence, timestamp ≈ 120 B) reach the backend over REST or WebSocket, so bandwidth stays flat no matter how long a session runs — and the video never exists outside the browser tab.

ARCHITECTURE FLOW

Data flow: laptop webcam to HTML5 getUserMedia to MediaPipe Pose Landmarker JS/WASM to kinematic state classifier to CodeIgniter 4 view and REST/WebSocket telemetry

Webcam → getUserMedia → Pose Landmarker (WASM/GPU) → kinematic classifier → UI overlay. Dashed paths are the only things that cross the network boundary.

KINEMATICS & MATHEMATICAL THRESHOLDS

MOVEMENT SIGNAL & FORMULA THRESHOLD EDGE HANDLING
Jumping Hip Δy (23/24) + ankle Δy (27/28)
v = Δy / 5
v > 0.045 · body_height Feet cut off by the webcam frame → hip-only fallback (Δy > 6% torso length)
Crouching Hip–knee–ankle interior angle at the knee
θ = angle(hip−knee, ankle−knee)
θ < 120° + hip below calibrated baseline Baseline hip position calibrated from 60 standing frames (median)
Kicking Ankle radial displacement from pelvic midpoint
d = |ankle_x − pelvis_x| / shoulder_width
d > 1.15 Anchor-leg stability verified ≥ 4 frames (rejects walk cycles)
Lying down Torso vector, shoulder midpoint → hip midpoint
φ = atan2(t.x, −t.y)
|φ| > 60° for ≥ 10 frames Also triggers when the whole body drops into the lower third of frame

MOVING AVERAGE FIFO

5–10 frame FIFO buffer per landmark, per axis. Removes single-frame landmark jitter (~1–2 px at 720p) before any threshold is evaluated.

DEBOUNCE 300 MS

State-lock after a detection fires, plus hysteresis (enter 0.045 / exit 0.030) so landing does not flicker between JUMP and CROUCH.

RUNTIME BUDGET

Inference 12–22 ms/frame, classifier < 2 ms, WASM heap ~30 MB. Fits the 33 ms frame budget at 30 FPS with room for rendering.

⬇ Download Architecture Specification (PDF)

2 pages · system flow, threshold table, state machine, performance budget

Certificates

divider

International Joint Lecture — Cybersecurity & Artificial Intelligence

Universitas Semarang (USM) × National United University (NUU), Taiwan • 7 October 2026 • 8 JP

Cybersecurity · Responsible AI & Cyber Resilience · AI Industry Collaboration · Generative AI & AI Companion Technologies

View certificate (PDF)

Certificate of Appreciation — International Joint Lecture USM × NUU Taiwan, 7 October 2026

Frontend Web Development

HTML, CSS, JavaScript, Responsive Design, Git

MySkill • Desember 2025

Frontend Certificate

Artificial Intelligence Specialization

AI/ML Concepts, Data to Deployment, AI Tools & Frameworks

MySkill • Desember 2025

AI Certificate

WordPress Development

CMS, Theme Customization, Plugins, Content Management

MySkill • Desember 2025

WordPress Certificate

Website Development Fundamental

Web Architecture, API, Cloud, Modern Development Tools

MySkill • Desember 2025

Web Development Certificate

Pelatihan Multimedia

Audio-Video Editing, Production, Client Collaboration

BLKK Semarang • Juli 2021

Multimedia Certificate

E-commerce Sales Growth

Increased sales from $250 to $3,000+/month

Luna Lumina • 2024-2025

Sales Achievement

Contact Me

divider
I'm open to collaboration, project discussions, or just connecting. Feel free to reach out!
Phone / WhatsApp +62 819 5238 7500
Address
Jl. Perbalan Timur, RT 03/RW 01
Mranggen, Demak, Jawa Tengah
Available for

Full-time, Freelance, Remote Work