Pravesh Nandanwar

Existing voice assistants are opaque, cloud-dependent, and treat user data as a product. There was no privacy-first assistant with genuine long-term memory and extensible tool use.

VoicePilot AI is a modular voice assistant with offline-first design. Wake-word triggers local speech recognition, which feeds into an LLM with a persistent vector memory store, enabling the assistant to recall context across sessions and execute tools on demand — all with explicit user permissions.

• Custom wake-word detection — activates without continuous microphone processing
• Offline-capable STT using Whisper for privacy-sensitive environments
• Long-term memory via vector embeddings — context survives restarts
• Extensible tool-calling framework — web search, calculator, system automation
• Local LLM support via Ollama, with cloud fallback for complex reasoning
• Explicit permission model — no action taken without instruction

The hardest problem was designing a memory system that's both fast enough for real-time conversation and persistent enough to recall context from weeks ago. The solution uses vector similarity for retrieval and a summarization layer to compress old context without losing meaning.

Manual resume screening is slow, inconsistent, and subject to human bias. Recruiters spend 70% of their time filtering before any meaningful evaluation begins.

TalentMatch AI automates the entire screening pipeline — parsing resumes, extracting skills, scoring against job descriptions, and ranking candidates using a hybrid NLP approach combining TF-IDF speed with BERT semantic accuracy.

• Batch resume parsing — PDF and DOCX with automatic text normalization
• Three matching strategies: TF-IDF (fast), BERT (accurate), Hybrid (production)
• ATS-style scoring with detailed per-criterion breakdown
• Intelligent skill extraction and missing skill gap analysis
• Interactive recruiter dashboard with visual analytics
• CSV and PDF report export for downstream processing

Personal trainers are expensive and not always accessible. People training alone lack real-time form feedback and objective rep counting, leading to inefficient workouts and injury risk.

PoseTrack AI uses your webcam to track 33 body landmarks in real time, analyzes joint angles, detects exercise phases, counts repetitions automatically, and streams the analysis to a responsive web dashboard — no hardware required.

• Real-time pose estimation at 24+ FPS using MediaPipe Pose
• Biomechanical joint angle models for 7+ exercises
• Automatic repetition counting via exercise phase detection
• Live calorie estimation based on rep rate and body weight input
• Workout history and analytics dashboard via Flask web app
• Works entirely in-browser — no app installation needed

Most personal finance apps are either too simple (spreadsheets) or too complex (enterprise tools). People lack a unified view of income, spending patterns, goals, and financial health in one place.

FinPilot is a production-deployed full-stack finance platform. It combines multi-account tracking, an ML-based Financial Health Score, anomaly detection, savings goals, and interactive analytics into a polished web application backed by Django REST Framework and PostgreSQL.

• Multi-account income and expense management with categories
• Financial Health Score computed from income, expense, and liability ratios
• Transaction anomaly detection and budget overspend alerts
• Savings goal tracking with dynamic progress visualization
• Spending trend forecasting using rolling window analysis
• Dark/light mode responsive UI
• Live production deployment on Render with PostgreSQL

Modern teams and individuals need a real-time communication platform that combines instant messaging, group collaboration, media sharing, and AI assistance — all in one scalable, production-ready system.

ChatFlow is a full-stack real-time messaging platform inspired by WhatsApp, Telegram, Discord, and Slack. It delivers instant messaging with Socket.IO, group conversations, media sharing via Cloudinary, AI-powered assistance, real-time notifications, typing indicators, read receipts, and an admin dashboard for platform analytics.

• Real-time messaging with Socket.IO — one-to-one and group conversations
• Message editing, deletion, emoji reactions, typing indicators, and read receipts
• Online presence detection and real-time notifications
• Friend requests, group creation, member roles, and conversation history
• Media sharing — image, video, and document uploads with Cloudinary storage
• AI-powered assistant for smarter communication
• JWT authentication with refresh tokens and bcrypt password hashing
• Admin dashboard with user management, analytics, and audit logs
• Responsive design with Dark Mode and glassmorphism UI
• Dockerized deployment with GitHub Actions CI/CD

The hardest challenge was implementing low-latency real-time communication while maintaining message persistence and delivery guarantees. The solution uses Socket.IO for WebSocket connections with fallback to polling, Redis for session management, and PostgreSQL with Prisma for reliable message storage — ensuring messages are never lost even during connection drops.

Job seekers struggle to find relevant opportunities, track applications, and gain meaningful career insights due to fragmented platforms and poor user experiences.

JobInsightPortal is a responsive frontend platform that provides an intuitive interface for exploring job opportunities, gaining career insights, visualizing job trends, and tracking applications — all with fast performance and seamless navigation across all devices.

• Fully responsive design — works seamlessly on desktop, tablet, and mobile devices
• Intuitive job search interface with advanced filtering
• Career insights dashboard with visual representation of job trends and statistics
• Application tracking system for managing job applications
• Smooth navigation with quick access to all sections
• Optimized for speed and accessibility
• Contact and inquiry forms for recruiter communication
• Article and resources section for career guidance

The main challenge was creating a fully responsive interface that works flawlessly across all screen sizes while maintaining fast load times. The solution uses mobile-first CSS with flexbox/grid layouts, optimized images, and lazy loading techniques to ensure smooth performance even on slower connections.

Financial institutions need secure, scalable banking platforms that can handle customer management, transactions, ATM operations, and content management — all while maintaining strict security standards.

JPCB Bank is a full-stack banking platform built with CodeIgniter 4 PHP framework. It features a comprehensive admin dashboard, integrated CMS for content management, ATM location and status tracking, secure transaction processing, and customer management — providing a complete banking solution.

• Complete admin dashboard for platform management
• CMS integration for dynamic content management
• ATM location tracking and status monitoring
• Customer account management and transaction processing
• Secure authentication and authorization system
• Responsive admin panel interface
• Database migration and seeding support
• Environment configuration for different deployment stages
• Composer-based dependency management
• MVC architecture following CodeIgniter 4 standards

The primary challenge was building a secure banking platform that balances functionality with strict security requirements. The solution implements CodeIgniter 4's built-in security features including CSRF protection, XSS filtering, and query binding for SQL injection prevention, along with role-based access control for the admin panel.

Field service businesses (HVAC, plumbing, electrical, landscaping) lack affordable, purpose-built CRM tools. They rely on spreadsheets, WhatsApp, and paper — leading to missed jobs, poor communication, no analytics, and no centralized customer management.

ServiceCRM is a full multi-tenant SaaS CRM built with PHP and MySQL. It features role-based access control (RBAC) with Admin and Super Admin panels, job dispatch, technician tracking, customer portal, automated invoicing, and comprehensive analytics — all in a secure, production-grade platform.

• Multi-Tenant Architecture — Complete data isolation per business with database-level separation
• RBAC (Role-Based Access Control) — Granular permissions for Super Admin, Admin, Manager, Technician, and Customer roles
• Super Admin Panel — Platform-wide oversight, tenant management, system settings, and global analytics
• Admin Panel — Business-specific dashboard, user management, job dispatch, and reporting
• Job Dispatch Board — Real-time technician availability, smart routing, and status tracking
• Customer Portal — Job status tracking, service history, feedback submission, and invoice viewing
• Automated Invoicing — Dynamic invoice generation, PDF export, and payment tracking
• Analytics Dashboard — Revenue metrics, job completion rates, technician performance, and customer satisfaction scores
• Real-time Notifications — Email and in-app alerts for job assignments, status updates, and reminders
• Secure Authentication — Password hashing, session management, and CSRF protection

The biggest challenge was implementing true multi-tenancy with complete data isolation while maintaining performance and scalability. The solution uses a database-per-tenant approach with shared lookup tables, combined with a robust RBAC system that ensures users from different organizations never access each other's data — all while delivering sub-200ms response times for complex dashboard queries.