Problem

Adoption, sales, lost and found, and pet services needed one account system, one set of APIs, and reliable payments.

What I built

• Built REST APIs and customer pages for adoption, pet sales, lost and found, breeds, and service discovery for 5,000+ accounts.
• Developed pet profile, diary, search, and account models across MongoDB and MySQL. Redis caching cut query processing time by 35% and API latency by 120 ms.
• Implemented Razorpay checkout, Cognito and JWT sessions, and Socket.io updates for messaging and live pet location.
• Stored pet media in Amazon S3 and purged cached files through the Bunny CDN API, with XML sitemaps, canonical URLs, and Schema.org structured data.
• Deployed the Node.js services with Docker, PM2, and GitHub Actions, and used Swagger docs for Postman checks. Public and auth routes use Helmet, rate limiting, and bcrypt.
• Connected OpenAI content tools and Firebase Cloud Messaging for pet profiles and customer activity.

Problem

Each branch was tracking stock, purchase orders, and bills separately. Wallets, pay-later accounts, supplier ledgers, and GST needed to sit in one portal.

What I built

• Built screens for pet sales, products, stock, purchase orders, sales invoices, returns, and supplier records.
• Implemented billing for customer wallets, pay-later accounts, supplier ledgers, and GST calculations.
• Added revenue charts, PDF invoices, and QR scanning for point-of-sale inventory intake.

Problem

Content, vendor onboarding, and support tickets were spread across the team, and every role could see too much.

What I built

• Built admin screens for pet breeds, vaccinations, pet names, NGOs, media, SEO metadata, and Schema.org.
• Developed vendor registration, QR tracking, user administration, and support-ticket workflows on the API.
• Enforced role-based access so distributors, operations staff, and support agents see only permitted actions.

Problem

Orders could go through without a balance check, and duplicate payments were hard to catch.

What I built

• Built order APIs with balance checks, account validation, and duplicate-payment protection for 2,000+ users.
• Added payment-gateway webhooks with checksum verification, reducing transaction discrepancies by 25%.

Problem

Classes, uploads, and grades were coordinated by hand.

What I built

• Built role-based access for classes, assignment uploads, and grading across 500+ students.
• Automated faculty reports and grade compilation, cutting manual coordination by 40%.

Problem

Historical crime records were hard to use for forecasting which crime types were likely, and where.

What I built

• Developed classification models (Random Forest, XGBoost) to predict crime categories.
• Engineered features from temporal and spatial data to improve prediction quality.
• Cleaned and preprocessed large-scale crime datasets.
• Visualized crime hotspots and trends using heatmaps and bar charts.

Problem

Insider privilege-escalation attempts are hard to separate from normal access. The models needed behavioural signals, not only static rules.

What I built

• Implemented ensemble models (LightGBM, Random Forest, AdaBoost) to classify threats with up to 97% accuracy.
• Used behavioural analytics and anomaly detection to identify escalation attempts.
• Trained models on a customized CERT dataset to simulate real-world scenarios.
• Proposed mitigation strategies based on access control policies.

Problem

Matching a face in a live video feed against known profiles had to work as images arrived, not only on saved photos.

What I built

• Built face detection and recognition modules using OpenCV and Haar cascades.
• Matched facial features against a database of known profiles.
• Integrated the system with video feeds for live tracking and alerts.

Problem

Lights, fans, and sensors needed one control path, and they stayed on when nobody was using them.

What I built

• Developed device control logic for lights, fans, and sensors using Java.
• Improved energy use by about 15% through scheduled device operations.
• Simulated IoT integration for a smart-home environment.