Data analyst / Python developer for small internal data apps (long-term, fully async)

We're a small insurtech company looking for someone reliable to maintain and occasionally extend a handful of small internal tools built with Python Streamlit. These apps are simple: they transform data files, generate documents, and produce reports around our core business. Nothing large or architecturally complex. What they need is regular, careful attention: small fixes, adjustments, occasional new tools from a clear spec. This is an ongoing, long-term collaboration, not a one-off gig. We want someone who gets to know our tools over time rather than starting from scratch each engagement. What you'll do Maintain and adjust small existing Streamlit apps (fixes, tweaks, small features) Occasionally build a new simple app from a written spec Above all, work carefully with data: transform it correctly, spot what looks off, and never break its meaning while reshaping it What we need from you You work fluently with AI tools and know how to get good results from them while keeping a critical eye on what they produce. This matters most to us: we want someone who uses AI well, not someone who avoids it or trusts it blindly. Strong data-handling skills: comfortable with Python, careful and rigorous with messy real-world data (mixed formats, missing values, inconsistencies) Able to build and maintain small Streamlit apps (the apps are simple, no heavy engineering needed) Comfortable with Excel files and APIs Basic clean Git workflow (branches, clear commits) Rigorous with testing: we care more about correct, well-verified output than fast delivery Fluent written English How we work Fully asynchronous. No standing meetings, no fixed hours. You work when it suits you. Volume is roughly 10 to 20 hours per month, ongoing. Default turnaround is 72 hours per task. Occasionally a task is flagged urgent and needs delivery within 24 hours, so we need someone who can be reactive when it matters. Communication and handoffs happen in writing, so clear written communication is essential. Selection process Shortlisted candidates complete a short paid test of about one hour. For the test, you'll record your screen while you work, with your voice, but no webcam and no camera. The point is for us to see how you actually work: how you approach the problem, and above all how you use AI tools along the way. Using AI is expected. What we want to observe is how you work with it, how you prompt it, and how you check and challenge what it gives you. To apply Please start your reply with the word "streamlit" so we know you've read the full posting. Then answer briefly: Share an example of data or app work you've done (a repo, a live app, or a short description of a real project). Tell us about a time you were transforming or cleaning a messy dataset and something didn't add up. What did you notice, and what did you do? Tell us about a specific time an AI tool gave you a wrong or misleading answer, and how you caught it