# Bagas Wastu - Full-stack developer specializing in AI-powered applications for startups and growing companies. ## About Me I'm a software developer who builds web applications with AI features that actually work. I help startups and growing companies automate repetitive tasks using practical AI integration. I've shipped several tools currently helping businesses automate research, generate content, and streamline workflows. ## Get In Touch - **Let's discuss your idea**: [Let's discuss your idea](https://cal.com/wastu/discuss) - **See my project**: [See my project](/projects) ## More About Me I'm a software developer who specializes in AI integration and web application development using React, Next.js, and modern AI APIs. I've built and deployed several AI features that are currently used in production applications. While I love collaborating within small teams, I often work as a solo developer on AI projects. This allows me to focus deeply on the technical challenges and deliver at my best pace. Whether you're looking to add AI capabilities to an existing product or build something completely new, I'd love to hear about your ideas. When I'm not coding, you'll find me reading a good book or manga, probably at one of Jakarta's coffee shops. I live in Jakarta (UTC+7) and keep flexible hours, which works well for collaborating with teams across different time zones. Have an AI project idea? I'd love to hear about it and discuss how we can bring it to life. ### Connect With Me - **X (Twitter)**: [@bgwastu](https://x.com/bgwastu) - **GitHub**: [bgwastu](https://github.com/bgwastu) - **LinkedIn**: [wastu](https://linkedin.com/in/wastu) - **Email**: [bagas@wastu.net](mailto:bagas@wastu.net) ## Projects Here are some of my featured projects: ### MAIA **Custom AI workspace platform for teams and communities** **Client**: Mayar.id **Year**: 2025 For six months, I worked full-time with the team at Mayar.id to help build their AI product, MAIA . The goal was ambitious: to create a platform that could compete with leading AI companies by offering advanced AI tools that actually do work for users, not just answer questions. ### The Problem We Solved Most AI platforms only let you chat with AI. But businesses need AI that can actually complete tasks, like researching competitors, gathering data from websites, or creating detailed reports. We wanted to build AI tools that could handle these jobs automatically, so teams could focus on bigger decisions instead of busy work. ### What I Built - **Deep Research**: An AI tool that could investigate any topic and gather comprehensive information from across the web. - **MAIA Browser Operator**: This tool could actively browse websites to find specific, real-time information, like pricing details or contact information. - **PDF Paper & Web Report Generator**: This took all the information gathered by the other tools and automatically organized it into clean, professional reports. ### The Result These features are now a core part of the MAIA platform, used daily by businesses across various industries. Tasks that used to take hours of manual research and compilation can now be completed in just a few minutes. It was incredibly rewarding to see these tools go live and make a real difference in how people work. --- ### Intervey **AI-powered survey platform that just feels like an interview** **Client**: Proof of Concept **Year**: 2024 I spent months talking to managers, analysts, and market researchers about their survey tools. Same complaint every time: "We get surface-level answers that don't tell us anything useful." So I built Intervey as a proof of concept to see if AI could interview people better than static questionnaires. The idea was simple. What if the AI could actually listen and ask follow-up questions when someone gives a vague answer? Dig deeper when they mention something interesting? ### What I learned building this The technical part was interesting but not the hard part. Getting the AI to extract meaningful insights from messy, emotional interview data? That was the challenge. I learned a ton about analyzing open-ended questionnaires at scale. When you have hundreds of responses with real feelings and nuanced opinions, you can't just throw them at an AI and expect magic. You need to understand context, read between the lines, build systems that catch contradictions and dig deeper. The trickiest part was generating useful reports and visualizations from raw emotional data. How do you turn "I feel frustrated sometimes but it's not terrible" into actionable insights? That took a lot of iteration. ### How it works The AI generates interview questions based on what you're trying to research. Then it actually conducts the interview, adapting based on each response. If someone gives a vague answer, it asks for specifics. If they mention something interesting, it follows up. After the interview, it analyzes everything and generates reports. No manual coding, no hours of reading through responses. Just insights. ### Where it stands now This project is shelved. No ETA on if I'll pick it back up. It proved the concept worked, but turning it into something people would actually pay for is a different problem. Still, I use what I learned here in almost every client project. Understanding how to make AI feel conversational instead of robotic? That's valuable everywhere. --- ### TesIELTS **AI-powered IELTS speaking test preparation** **Client**: SaaS **Year**: 2024-2025 Back in 2024, Faiz reached out to me about a problem he kept seeing. People in Indonesia wanted to study or work abroad, but the IELTS exam was expensive and intimidating. The hardest part? Practicing speaking. You could hire an expensive tutor or talk to yourself in your room, never knowing if you were improving. Faiz had the English teaching and IELTS expertise. I had the technical skills. So we built TesIELTS together. This was my first real SaaS with actual paying users. Real people depending on it to prepare for life-changing exams. ### The Problem I Wanted to Solve There weren't many good ways to practice IELTS speaking without spending a fortune. You could hire an expensive tutor or practice alone in your room, never knowing if you were actually getting better. I wanted to build something that could give the same detailed feedback as a real IELTS examiner, but at a price students could afford. ### My Solution: An AI That Acts Like a Real Examiner Faiz and I built TesIELTS as an AI coach that works just like the real test. Students have actual conversations with the AI, and then get detailed feedback on four key areas: The AI listens to how you speak. It catches when you hesitate, checks your grammar, scores your pronunciation. Students get band scores (1-9) for each area, plus specific tips on what to improve, exactly like a real IELTS examiner would give. - **Fluency & Coherence**: How smooth and logical your speech sounds - **Vocabulary**: The variety and quality of words you use - **Grammar**: How correct and complex your sentences are - **Pronunciation**: How clear and natural you sound ### The Technical Challenge The hardest part was teaching the AI to listen like a human examiner. The system had to understand how students said things, catch tiny pauses, judge grammar complexity, and score pronunciation quality. This meant building advanced voice analysis technology that goes way beyond basic speech recognition. ### The Result The platform now helps over 100 students every month prepare for their exams. After building and growing it successfully, I completed my exit from the company. Looking back, the most valuable lesson was learning how to build AI that evaluates language skills as accurately as human examiners. Turns out you can make quality test prep available to anyone with an internet connection. You just need to actually understand what "quality" means first. --- ### Trendjacking **AI-powered marketing ideas generator based on current trends** **Client**: DOKI.id **Year**: 2024-2025 Narawastu from DOKI reached out with a problem. DOKI works with big brands like Danone Aqua, Fonterra, and Tokopedia. Their clients kept running into the same issue: by the time marketing teams found a trend and pitched ideas, the moment had passed. They needed something that could catch trends while they were still fresh, specifically on short-form video platforms. TikTok and Instagram Reels, where trends die in 48 hours. ### The technical challenges The hardest part was scraping TikTok and especially Instagram. Both platforms really don't want you doing that. Lots of reverse-engineering, proxy rotation, and dealing with rate limits. Instagram was particularly painful. Once I got the content, I needed the AI to actually watch the videos, analyze the stats, read the comments, and understand the sentiment. Then match all that data with corporate brand identities. The system had to know if a trend would actually work for Aqua's brand voice versus Tokopedia's. ### How it works I built an AI recommendation engine that analyzes everything automatically. Every piece of content gets tagged and meticulously analyzed. The AI watches the videos, checks engagement stats, and reads comment sentiment to understand the full context of each trend. Then it matches trends with specific brands. The system knows each client's voice and values, so it only suggests ideas that would actually make sense for them. ### The Result What used to take DOKI's creative team 1-5 days of research and brainstorming now takes 5 minutes. Their clients can respond to trends while they're still trending, which is the whole point. The platform is now part of DOKI's core workflow for their major clients. Watching it go from concept to daily-use tool for big brands was pretty satisfying. --- ### Verbata **AI tool that codes thousands of interview responses in minutes** **Client**: Deka Insight **Year**: 2024 Deka Insight, one of Indonesia's most respected market research companies, had a problem. Their researchers were spending 1-3 days manually reading and categorizing thousands of interview responses - a process called "verbatim coding." Tedious work that kept their best minds tied up instead of doing actual analysis. ### The MVP I built the first version fast. With a simple interface, one job: read text, categorize it, done. No fancy features, just the core functionality they desperately needed. The goal was to prove the concept worked before investing more time. It worked. What took researchers days now took 5 minutes. But using it revealed what was missing - the researchers needed more control over how the AI interpreted their data. ### Version 2: Making It Actually Good I rebuilt it from scratch with a blue theme and much more robust features: - **Customizable prompts** - researchers can tweak how the AI interprets responses - **Better coding process** - more transparent about decision-making with confidence scores - **Batch processing** - handle hundreds of responses at once - **Quality checks** - human-in-the-loop verification to maintain research standards ### The Hard Parts Making the AI accurate enough for ISO-certified standards while handling Indonesian language nuances and research-specific terminology. Also had to keep everything secure since they're dealing with sensitive client data. The trickiest part was understanding the subtle emotional context that makes qualitative research valuable - you can't just throw text at an AI and expect it to get the nuance right without proper framing. ### What I Learned Ship the MVP fast, get real feedback, then rebuild properly. The first version proved the concept worked. The second version made it actually useful for their daily workflow. Now they spend less time on manual data entry and more time finding insights. --- ## Experiments A collection of small ideas and weekend projects brought to life out of curiosity. ### Parsley **AI document parser that transforms PDFs or images into structured JSON or CSV data** **Year**: 2025 I used to build custom OCR systems for every document type I needed to parse. Bank statements were especially tedious: hardcoded pixel positions, regex patterns for each bank's format, and brittle parsing logic that broke with every minor PDF template change. Then I'd need similar setups for invoices, receipts, forms. ## The solution Rather than keep fighting with OCR, I built Parsley with LLMs. It understands the content directly and structures it as needed. If you ask for "customer name" it finds it, whether it's labeled "Bill To:", "Customer:", or hidden in a paragraph. No hardcoded positions or regex. Your API keys are used directly in the web app, so documents go straight to Google or OpenRouter. I can't access them. Everything stays stateless. ## Features - Custom schemas (define your own structure, or let AI generate it) - Supports PDF (including password-protected) and images (PNG, JPEG, WebP) - Multiple AI providers (Google Gemini, OpenRouter, your own keys) - Demo mode with rate-limited free tier (no API key needed) - Export as JSON or CSV - API works with n8n, Zapier, or other automation tools ## How I use it I run invoices through Parsley in n8n, extract the needed data, and send it straight to my accounting spreadsheet. Same approach for bank statements, receipts, forms, any document where I want structured data fast. --- ### Notestorm **Minimalist writing app with AI that keeps you in flow** **Year**: 2025 I built this scratchpad for those frustrating moments when I know what I want to say, but the words just won't come out. By the time I find the right words, I've completely lost my train of thought. ## What makes it different Most note apps interrupt my flow. Notestorm keeps me writing by suggesting completions that match how I sound. I can skip the words I'm stuck on and keep the ideas flowing. Everything runs locally, my notes never leave my device. ## Features - AI autocomplete that learns your writing style - Multiple AI provider support (Google, Groq, Anthropic, OpenAI, OpenRouter) - VS Code keybindings support - Perfect for brainstorming, drafting emails, or scratch notes - Optional Chrome built-in AI support (Canary only) ## How I use it Quick brainstorming sessions, drafting emails before copying to Gmail, and temporary notes I know I'll delete later. ## Building it Spent 5 days building this with TanStack Start, CodeMirror 6, and Vercel AI SDK. Getting the autocomplete UX right is the hardest part. When to trigger suggestions, which keyboard shortcuts feel natural, and keeping it instant. --- ### DeleteX **Selectively delete your content on X (formerly Twitter)** **Year**: 2024 I wanted to clean up my X timeline but didn't want to nuke everything. So I built this tool to selectively delete tweets, retweets, and likes based on whatever criteria I need. ## How it works DeleteX uses your [X archive data](https://help.x.com/en/managing-your-account/how-to-download-your-x-archive) to generate a userscript that runs in your browser. The script only deletes what you selected. Everything happens locally in your browser, so your data never leaves your device. ## Tech stack Built with Next.js and [PGLite](https://pglite.dev) for the database. PGLite also powers the search functionality using PostgreSQL full-text search, making it fast to filter through thousands of tweets. --- ### Cek Sandi **Password strength checker using zxcvbn algorithm** **Year**: 2024 I built this password strength checker to give awareness to people and developers about proper password rules. The goal? So we don't end up with more entries on [dumbpasswordrules.com](https://dumbpasswordrules.com/). ## What makes it different Most password checkers are simplistic. Uppercase, lowercase, numbers, symbols, done. Cek Sandi uses the [zxcvbn algorithm](https://www.usenix.org/conference/usenixsecurity16/technical-sessions/presentation/wheeler), the same tool security professionals use. It analyzes common words, keyboard patterns, and predictable substitutions to show how resistant your password actually is to cracking attempts. ## Features - Accurate strength analysis using zxcvbn algorithm - Tips for creating strong passwords (passphrases, not just random characters) - Open source, you can inspect the code and verify its security - Focused on Indonesian users with localized tips Built with Next.js and Mantine, deployed on Cloudflare Pages. --- ### Moonlit **Slowed/nightcore effects for your favorite YouTube & TikTok videos** **Year**: 2023 I'm kind of a weird person who can only focus when listening to nightcore or slowed+reverb looped music. So I built a music player that lets me customize playback speed and reverb in real-time. ## Features - Change playback speed and reverb effect in real-time - 3 default modes (slowed, normal, speed up) with a customizable mode - Custom background for personalized aesthetics - YouTube integration with quick link (youtubelit.com) ## The backstory I used to reverse-engineer my audio driver just to access the legacy Realtek HD Audio Manager. It was the only way to change output audio pitch manually. I even wrote [a simple script](https://github.com/bgwastu/ytnc_cli) to convert YouTube songs into nightcore, but it was painfully slow and not very customizable. At first, I was very reliant on AudioContext API for everything. But then I realized I could just use playback speed on the video element and only use AudioContext for reverb. Much simpler. Building this taught me how audio and frequencies work, and more importantly, not to waste time writing features that nobody (including myself) would actually use. --- ### XY Puzzle **Escape room-style puzzle for tech enthusiasts** **Year**: 2023 Five days before TeknumConf 2023, I had a random idea: what if my name card was a puzzle? Medium complexity escape room-style challenge that requires some cryptography knowledge, but nothing too deep. ## The story I wanted something like [hacker.gifts](https://frantic.im/hacker-gifts/) but more accessible. Had three days to pull it off - one day for designing and ordering the card, two days for coding the puzzle using Next.js server components (which was brand new tech at the time, making things way harder than expected). ## The plot twist After distributing the cards at the conference, no one could solve it. Maybe it was too hard, or maybe they just weren't motivated enough. So I recently updated the puzzle to be more accessible while keeping it challenging. Building this pushed me to think creatively and gave me hands-on experience with the app router and server components. Try [the puzzle](https://xy.wastu.net) yourself, even without the name card. --- ## Metadata - **Total Projects**: 5 - **Total Experiments**: 6 - **Last Updated**: 2026-05-24 - **Generated for**: AI crawlers and search engines