01/03/2026
𝐑𝐞𝐝𝐞𝐟𝐢𝐧𝐢𝐧𝐠 𝐏𝐫𝐨𝐩𝐓𝐞𝐜𝐡 𝐰𝐢𝐭𝐡 𝐀𝐈: 𝐓𝐡𝐫𝐢𝐥𝐥𝐞𝐝 𝐭𝐨 𝐮𝐧𝐯𝐞𝐢𝐥 𝐦𝐲 𝐥𝐚𝐭𝐞𝐬𝐭 𝐟𝐮𝐥𝐥-𝐬𝐭𝐚𝐜𝐤 𝐃𝐣𝐚𝐧𝐠𝐨 𝐩𝐫𝐨𝐣𝐞𝐜𝐭, 𝐍𝐢𝐫𝐚𝐩𝐨𝐝𝐍𝐞𝐬𝐭!
I wanted to build more than just a real estate listing site. I wanted to engineer a secure, intelligent ecosystem that actively works for its users. By leveraging a custom dataset of over 32,000 property records, I integrated advanced Machine Learning architectures and strict security protocols to completely eliminate brokers and create total market transparency.
Here is a look under the hood at the engineering behind the NirapodNest system:
𝐓𝐡𝐞 𝐌𝐚𝐜𝐡𝐢𝐧𝐞 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐄𝐧𝐠𝐢𝐧𝐞 (𝐓𝐫𝐚𝐢𝐧𝐞𝐝 𝐨𝐧 𝟑𝟐𝐤+ 𝐑𝐞𝐜𝐨𝐫𝐝𝐬)
𝐏𝐫𝐢𝐜𝐞 𝐄𝐬𝐭𝐢𝐦𝐚𝐭𝐨𝐫 & 𝐌𝐚𝐫𝐤𝐞𝐭 𝐕𝐚𝐫𝐢𝐚𝐧𝐜𝐞 (𝐗𝐆𝐁𝐨𝐨𝐬𝐭 𝐑𝐞𝐠𝐫𝐞𝐬𝐬𝐨𝐫): A custom XGBoost model analyzes live market data to instantly calculate an accurate predicted asking price for sales or rentals. To empower users, the system automatically compares the owner's listed price against the AI prediction, displaying the exact market variance and highlighting any markup in real-time.
𝐍𝐚𝐭𝐮𝐫𝐚𝐥 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐒𝐞𝐚𝐫𝐜𝐡 (𝐬𝐩𝐚𝐂𝐲 𝐍𝐋𝐏 + 𝐑𝐞𝐠𝐞𝐱): To make the platform hyper-accessible, I implemented a smart text-based search. Using spaCy for NLP combined with custom Regex parsing, users can simply type their requirements in plain English, and the engine instantly translates their natural text into complex database queries.
𝐒𝐮𝐩𝐩𝐨𝐫𝐭 𝐂𝐡𝐚𝐭𝐛𝐨𝐭 (𝐑𝐀𝐆 𝐯𝐢𝐚 𝐅𝐀𝐈𝐒𝐒): Built a Retrieval-Augmented Generation (RAG) pipeline using LangChain and a FAISS vector database to answer complex user FAQs instantly based on semantic meaning, bypassing the limitations of basic keyword matching.
𝐍𝐞𝐢𝐠𝐡𝐛𝐨𝐫𝐡𝐨𝐨𝐝 𝐈𝐧𝐬𝐢𝐠𝐡𝐭𝐬 (𝐆𝐞𝐨𝐬𝐩𝐚𝐭𝐢𝐚𝐥 𝐋𝐨𝐠𝐢𝐜 𝐄𝐧𝐠𝐢𝐧𝐞): A custom geospatial engine that goes beyond standard amenities. It generates comprehensive neighborhood insights and assigns a custom livability rating based on specific property details and location data, helping users understand the true vibe of an area.
𝐒𝐦𝐚𝐫𝐭 𝐀𝐮𝐭𝐨-𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐨𝐫: An NLP tool that cross-references a property’s specs and location to automatically write highly professional, context-aware property descriptions for owners.
𝐌𝐚𝐫𝐤𝐞𝐭 𝐑𝐚𝐝𝐚𝐫 𝐀𝐥𝐞𝐫𝐭𝐬 (𝐊-𝐍𝐞𝐚𝐫𝐞𝐬𝐭 𝐍𝐞𝐢𝐠𝐡𝐛𝐨𝐫𝐬): Users don't have to endlessly scroll. I engineered a highly customizable Market Radar using a K-Nearest Neighbors (KNN) algorithm that constantly scans the database for properties matching a user's exact multi-dimensional criteria. It pushes alerts via the in-app notification bar and email based on user preference: Instant, Weekly, or at a Daily User-Defined Time.
𝐌𝐮𝐥𝐭𝐢-𝐋𝐚𝐲𝐞𝐫 𝐒𝐞𝐜𝐮𝐫𝐢𝐭𝐲 & 𝐙𝐞𝐫𝐨-𝐓𝐨𝐥𝐞𝐫𝐚𝐧𝐜𝐞 𝐅𝐫𝐚𝐮𝐝 𝐃𝐞𝐭𝐞𝐜𝐭𝐢𝐨𝐧: Security is non-negotiable. I implemented a strict, multi-layer verification process during account creation. If a new user attempts to register using the sensitive details (like NID or a claimed property address) of an already verified user, the system's anti-fraud engine catches it and executes an immediate, automatic ban on the fraudulent account.
𝐑𝐞𝐚𝐥-𝐓𝐢𝐦𝐞 𝐂𝐨𝐦𝐦𝐮𝐧𝐢𝐜𝐚𝐭𝐢𝐨𝐧: To completely remove the middleman, I built a seamless, real-time messaging architecture allowing verified buyers and renters to chat directly with property owners.
𝐂𝐨𝐫𝐞 𝐒𝐭𝐚𝐜𝐤: Python, Django, PostgreSQL, XGBoost, Scikit-Learn, spaCy, LangChain, FAISS, React JS.
Building NirapodNest pushed me to bridge the gap between heavy AI data processing, strict backend security, and a flawless user experience.
Check out the video below to see the AI and security features in action! I’d love to hear your thoughts or feedback in the comments.
