techwizard

Deploying DocuGPT (ChatPDF) to the Cloud

Status: Live & Stable Stack: Flask, MongoDB, Pinecone, LangChain, Hugging Face Spaces (Docker)

1. Overcoming RAM Limits My first attempt on Render failed instantly. The app crashed with an “Out of Memory” error because my AI library (sentence-transformers) needs ~800MB RAM, but Render’s free tier only provides 512MB.

2. The Move to Hugging Face I switched to Hugging Face Spaces using the Docker SDK. This was the critical fix—it offers 16GB of RAM for free, allowing the heavy AI models to run smoothly without crashing.

3. Git & Auth Fixes I hit two walls: a “refusing to merge unrelated histories” error and rejected password authentication. I fixed the merge by forcing it with the –allow-unrelated-histories flag. For auth, I generated a Write-Scope Access Token and injected it directly into the git remote URL to bypass the deprecated password prompt.

4. Stabilizing the Server The app was crashing during PDF uploads with a [CRITICAL] WORKER TIMEOUT. Gunicorn was killing any process that took longer than 30 seconds. I updated the Docker configuration to allow a 5-minute timeout, ensuring large PDFs process fully.

5. Final Polish I fixed a login bug caused by iframe cookie blocking by adding a direct access link, and I upgraded the UI with a modern Tailwind gradient design.

Outcome: DocuGPT is now live, stable, and running on a robust 16GB environment.

Ever finished a movie and wondered, “What should I watch next?” I spent this week building a Content-Based Recommendation Engine to answer exactly that.
Here is how I built it day-by-day.
📅 Day 1: The Setup
I skipped the classic MovieLens dataset (ratings-based) and chose the TMDB 5000 Dataset to focus on content analysis. Using Pandas, I merged the movie and credits files and filtered for the essentials: genres, keywords, overview, cast, and crew.

📅 Day 2: Wrangling Data
Data cleaning was the heavy lifting. Columns like genres were stored as JSON strings (e.g., [{"id": 28, "name": "Action"}]).

The Fix: Used ast.literal_eval to parse them into Python lists.
Feature Engineering: I extracted the top 3 actors and the director. I also collapsed spaces (e.g., “Science Fiction” → “sciencefiction”) to create unique tag entities.
Result:A single “Super Column” called tags that summarizes the entire movie.
📅 Day 3: The Math (Vectorization)
To measure similarity, I needed to turn text into numbers.
Vectorization: Used Scikit-Learn’s CountVectorizer (Bag of Words) to convert tags into 5,000-dimensional vectors, removing stop words.
Similarity:Used Cosine Similarity to measure the angle between vectors. This generated a matrix comparing every movie against every other movie.
📅 Day 4: The Interface
I used Streamlit to build a frontend.
Logic: The user selects a movie → App finds its index → Sorts the similarity matrix → Returns the top 5 matches.
📅 Day 5: API Integration
Text-only lists are boring. I signed up for the TMDB API and wrote a script to fetch real-time movie posters. Displaying them side-by-side made the app feel like a real product.
📅 Day 6: Optimization
Re-calculating the model on every reload was too slow.
Solution: I used pickle to save the processed data and similarity matrix. The app now loads pre-computed files instantly.

How I Built a Car Price Predictor to Start My 2026 Dev Journey

Date: January 1, 2026 By: Hardik Mittal

They say the best way to learn is to build. So, while everyone was making New Year’s resolutions, I decided to open my IDE and ship a project.

Here is the story behind my latest build: a Car Price Predictor.

1. The Problem 🧐

Selling a used car is confusing. Prices vary wildly based on fuel type, kilometers driven, and the age of the car. I wanted to build a tool that uses actual data to give users a fair price estimate instantly.

1. The Machine Learning Model 🤖

I started with a dataset of utilized cars. The data cleaning phase was crucial—handling missing values and cleaning up inconsistent car names. I used Scikit-Learn to train a Linear Regression model. To handle the categorical data (like “Petrol” vs “Diesel” or “Maruti” vs “Hyundai”), I implemented a pipeline using OneHotEncoder. The model achieved an accuracy (R² score) of approx 0.84, which is solid for real-world estimation.

1. Building the Web App 💻

A model inside a Jupyter Notebook isn’t useful to anyone. I needed a frontend.

Flask was my choice for the backend API.

For the frontend, I wanted a clean, “tech” aesthetic. I built a responsive form using CSS Variables for easy theming (and a dark-mode-ready structure).

JavaScript handles the logic. I created dynamic dropdowns so that when a user selects “Honda,” the model list only shows Honda cars.

1. The Result 🎉
   The app is now live! You can input your car’s details, click “Predict,” and get a price estimate in milliseconds without refreshing the page.

2. What’s Next?
   I plan to deploy this soon and maybe add more features like current market trends graphs.
   Happy New Year and Happy Coding!

I’m excited to share that I’ve successfully deployed my latest Machine Learning project: an SMS & Email Spam Detector.

💡 The Problem: We all get annoying spam messages. I wanted to build a model that could filter them out in real-time using Natural Language Processing (NLP).

🛠️ The Tech Stack:

Python & Scikit-Learn: For building the model (tested Naive Bayes, SVM, and Voting Classifiers).

NLTK: For text preprocessing (tokenization, stemming, stopword removal).

Streamlit: For creating the interactive web interface.

Render: For cloud deployment.

📉 The Challenges: It wasn’t a straight path! I spent hours debugging NotFittedErrors, fixing sparse vs. dense matrix mismatches, and resolving dependency conflicts during deployment. But getting that green “Deploy Succeeded” checkmark was worth it.
#MachineLearning #Python #Streamlit #NLP #DataScience #OpenSource #Coding

We’ve all stared at a blank Instagram or LinkedIn profile, wondering what to write. I wanted to build a tool that solves “writer’s block” for social media. The goal was simple: Input a topic, get a creative bio or trending hashtags instantly.

I’m working on my first project! This is so exciting. I can’t wait to share more updates as I build. I create a portfolio.