Web Scraping FAA.gov
About
First UpWork contract. Scraped FAA.gov aviation data, cleaned and delivered in CSV format. Review (5 Stars): Tikita did the task exactly as I wanted and she did it quickly. She communicated well and our intermediate check ins were helpful. Would definitely hire again.
Tech Stack
Links
Discussion & Feedback
Have questions about this project? Built something similar? Share your thoughts!
(Requires GitHub account to comment)
Related Projects
Stardew Valley Wiki Chatbot GPT
Knows everything about Stardew Valley! Ask any Stardew Valley question and get instant answers - temporarily offline after OpenAI took it down. Reached 1K chats and 10 reviews (4.2 star rating) before going down. Built with Puppeteer and Cheerio for scraping the Stardew Valley Wiki, TypeScript for structuring data into markdown files, and sklearn for clustering content into 9 categories to speed up responses.
Hackathon - UoM StudentHack
I got to take part in a Hackathon with UniCSHackathons! It was an awesome experience and my team were amazing to work with - we created a site where you can choose between 3 versions of real news articles, aimed at primary, secondary and adult ages to learn about current space events! The articles were scraped using puppeteer/cheerio from spacenews.com then stored on a PostgreSQL database using a Django backend, before being sent to Google's Gemini LLM to create the new content. The new articles were displayed with a React and typescript frontend!
Brand System
## Voice Profile System Built LinkedIn voice DNA profiles for both Daeda founders by ranking the top 10 posts each (weighted reactions × engagement rate) and reverse-engineering the patterns into reusable style guides the AI curation pipeline can draft against. ### Outputs - **tikita-voice.md** - short, punchy fragments, problem-first hooks, British-casual builder-sharing-workshop tone, moderate emoji, Unicode bold for emphasis. Average post 150-300 words. - **jack-voice.md** - long-form thesis-argument-conclusion (300-500+ words), provocative claim hooks, tab-indented numbered lists, gaming and fantasy metaphors, almost no emoji or hashtags. - Side-by-side comparison table so the right voice can be picked consciously when drafting. - Source data preserved as CSV per founder per year. ### Why It Matters Tone is the hardest thing to keep consistent across an AI writing pipeline. By turning each founder's top performers into a documented style, the curation workflow drafts in either voice on command instead of producing generic LinkedIn-ese. ## Branded Asset Library A reusable library of branded assets - UI animation clips, motion templates, type treatments, colour pairings, product UI captures - aggregated to slot into the AI loop for video production. Why a library rather than per-asset generation: brand consistency at volume is hard. Pulling from a tagged, curated set means generated content can't drift off-brand. Currently used as the constraint layer in the daeda_tech YouTube AI loop.