🚀 From Streamlit App → Flask API → Full Deployment Earlier, I built a Diabetes Prediction app using Machine Learning and Streamlit. Now, I’ve taken it a step further by converting it into a complete backend + frontend system 🔥 🔹 Built ML model for prediction 🔹 Created Flask API for serving predictions 🔹 Tested endpoints using Postman 🔹 Deployed API on Render (Live) 🔹 Integrated with Streamlit UI This project helped me understand how ML models are actually deployed and used in real-world applications 💻 🔗 Live App: https://lnkd.in/dwBr79e5 🔗 API: https://lnkd.in/grwbgaXw #MachineLearning #Flask #Streamlit #API #Deployment #Python #DataScience

I always thought getting started with Machine Learning would be complicated. So instead of overthinking it, I tried building something small. I created a simple Android app that uses a machine learning model running on-device. Here’s what I did: • Trained a basic model in Python to learn a simple pattern (y = 2x) • Converted the model to TensorFlow Lite (.tflite) • Integrated it into an Android app • Built a small UI using Jetpack Compose to take input and show output Example: Input: 5 → Output: ~9.96 It’s a simple use case, but it helped me understand the full flow: data → training → conversion → mobile inference A few things that stood out: • ML models don’t give exact answers, they approximate • Running models on-device with TFLite is fast and practical • As an Android developer, getting started with ML integration is more approachable than it seems If you’ve been thinking about exploring ML but haven’t started yet, building something small like this is a good way to begin. It doesn’t need to be complex to be useful. Tech used: Python, TensorFlow, TensorFlow Lite, Kotlin, Jetpack Compose This was a small step, but a good starting point to understand how ML fits into mobile apps. #Android #MachineLearning #TensorFlowLite #JetpackCompose

Update: Smart Expense Tracker — Improved Version Since the first release, I have made several significant improvements to the app. What's new: Multi-user support with a login system Add, edit and delete expenses Improved automatic category prediction using ML Cleaner UI for better usability Persistent data storage using SQLite Fixed deployment and Docker configuration issues This project started as a simple idea but debugging and deploying it taught me a lot about real-world development — from ML model integration to database management and containerisation. Live App: https://lnkd.in/gGxsJ6tq GitHub: https://lnkd.in/gD9sA7ej #Python #Streamlit #MachineLearning #DataScience #Docker #HuggingFace #GitHub

Mini Project -🚦 Smart Traffic Violation Logger (Flask) Built a Flask-based web app to digitize traffic violation records with QR-enabled challans for quick status tracking at Cybernaut EdTech 🔍 Key Highlights: ▪️End-to-end CRUD operations using Flask and SQLAlchemy ▪️Dynamic QR code generation for real-time status tracking ▪️Searchable and filterable violation history ▪️Secure login system for authorized personnel ▪️Clean, responsive UI built with Bootstrap 💡 Demonstrates practical use of Flask, databases, and QR integration for smarter traffic management. #Flask #WebDevelopment #Python #Innovation

🚀 Built a small but useful project recently — a Desktop Notification & Reminder System. The idea was simple: don’t miss important tasks. So I created a Python-based app (Tkinter GUI) that lets users set reminders and get desktop notifications at the right time. What I worked on: – Designing a clean and simple interface – Handling time-based triggers – Making sure notifications are reliable It was a good hands-on experience with GUI development and practical problem-solving. Still improving it, but it already feels like something I’d actually use daily #Python #Projects #SoftwareDevelopment #Productivity #LearningJourney

Can Android and a laptop really transfer files over BLE? I built the PoC to find out. I wrote a short piece on Medium about building a BLE file transfer proof of concept between an Android app and a laptop acting as the peer. https://lnkd.in/dcdDJTbR The result: a working setup that could browse files, push uploads to Android, and pull files back out again, with a few useful failures along the way that made the final PoC much more credible. If you’re interested in Android, BLE, Python, or pragmatic prototyping, give it a read. #Android #Kotlin #AndroidDev #BLE #Python #MobileDevelopment #SoftwareEngineering #Prototyping #TechWriting #Medium

Built an AI Assistant… now turned it into a Web App using Streamlit 🚀 From: ➡️ PromptTemplate + LangChain To: ➡️ Interactive AI UI This blog shows how to take your AI project from code → real product. 📖 https://lnkd.in/grxN9kJq #AI #LangChain #Streamlit #Python #Developers #AIProjects #BuildInPublic #Tech #ArtificialIntelligence #Beginners #WebDevelopers

📱 Turning “wasting time on my phone” into something fun (and useful) We all do it. Pick up the phone → scroll → switch apps → repeat. So I tried something different. 🧠 What if your phone had a terminal + AI agent? I set up Termux + proot-distro to run a tiny Ubuntu environment. Not for serious work. Just for… fun experiments. 🧪 Tiny runtime = instant testing The fun part is you actually have real runtimes on your phone: Node.js Python So it’s not just “asking AI”. You can: generate a small script run it immediately tweak and re-run in seconds. ⚡ Now instead of scrolling, I do this When I catch myself wasting time, I open terminal and: ask random technical questions explore weird ideas prompt AI to test something small run quick scripts just to see what happens Like: “what’s the simplest way to simulate this?” “write a tiny script to test this logic” And the agent just… does it. #termux #android #free-model #testing #poc #opencode #neovim

Quick FastAPI tip 💡 If you're still on @app.on_event("startup") / @app.on_event("shutdown"), those are deprecated. Move to lifespan. @asynccontextmanager async def lifespan(app: FastAPI):     app.state.db = await create_db_pool()     yield     await app.state.db.close() app = FastAPI(lifespan=lifespan) A few reasons I've come to prefer it: - setup and teardown live in the same function, so the lifecycle is obvious at a glance - cleanup still runs if something goes wrong - the same context manager could be used in tests as well - it's the recommended approach going forward Migrated a couple of services last week, took maybe 5 minutes each. Worth it #FastAPI #Python #Backend

Ten Shiny apps. One Electron shell. One very confused laptop. Quick dispatch from the Holy (native) Grail quest: last week we teased Shiny for Python running inside an Electron app on Windows. This week the temple doors creaked open a little wider, and we got the same Shiny app running in five different runtime modes at once, ten apps in total across R and Python, all wrapped in Electron and exported using one R package. The lineup: - R and Python Shiny, side by side - Native flavors: system, bundled, and auto-download - Containerized via Docker - Shinylive for more ephemeral work The dream is the boring one: you write a Shiny app, your non-technical users double-click an icon, and nobody has to learn what a port is. Different runtime modes for different deployment stories, same Shiny app underneath. No public-facing code yet. Parts of this have lived as scattered experiments on my GitHub for years: the current shinyelectron package only does shinylive export, and there was an earlier demo embedding a Docker container inside an Electron app. The new pieces, native auto-download, system, and bundled, are what finally tie it all together. More soon. Screenshot below shows all apps running at once.... #rstats #python #rshiny #pyshiny #electron #shinyelectron #datascience