Problem: Standard CMS platforms are often bloated and slow, requiring massive monolithic backends to serve simple content.
What I Built: A full-stack, decoupled blogging platform featuring a custom-built React CMS dashboard, secure JWT user authentication, and a headless REST API backend.
Result: A production-ready, highly-scalable platform capable of autonomously generating static SEO-friendly pages while restricting raw endpoints via custom middleware.
A rich text editor integrated seamlessly allowing for rapid blog post creation.
Server-side generated pages pulling directly from a custom NoSQL database implementation.
Encrypted user authentication restricting content creation to verified administrators.
Problem: Cybersecurity intelligence requires digesting massive amounts of raw log data in split seconds to respond to threats.
What I Built: A centralized React dashboard mapping simulated cyber intrusions to an interactive geographic UI for real-time threat intelligence.
Result: An immersive, command-center UI capable of digesting mock asynchronous WebSocket streams and mapping high-velocity telemetry onto actionable front-end components.
Mock intrusion alerts mapped in real-time onto an interactive geographical layout.
Simulated counter-measures panel allowing users to instantly block compromised IPs.
Detailed chronological history of all simulated network events and administrator responses.
Problem: Single-prompt LLMs fail at massive multi-step logical workloads that require external API access and tool routing.
What I Built: A dynamic multi-agent system utilizing LangChain to autonomously route and synthesize outputs from specialized Python worker scripts.
Result: Successfully engineered an autonomous "Supervisor" node traversing state graphs, validating parallel tool executions, and synthesizing modular results without human intervention.
Routes tasks intelligently to the most capable worker agent using an LLM-driven state machine.
Specialized nodes executing explicit Python scripts or API calls autonomously.
Global context injected into every prompt enabling multi-turn collaboration.
Problem: Rendering complex thousands-node relationship hierarchies causes significant UI lag when updating standard DOM elements.
What I Built: Programmed a high-performance JavaScript physics engine capable of dynamically rendering abstract data structures natively on the HTML5 Canvas.
Result: Achieved fluid 60fps graph interactivity leveraging custom mathematical friction, spring forces, and direct context rendering to bypass heavy framework bottlenecks.
Custom calculated friction and spring forces to organically place network nodes.
Drag, drop, and inspect nodes natively using HTML5 Canvas API context listeners.
Dynamic JSON ingestion scripts that instantly render graph nodes across the screen.
Problem: Generic fitness apps don't adapt to injury — they prescribe the same routine regardless of pain signals, leading to setbacks.
What I Built: A full-stack AI fitness companion that logs daily recovery scores, tracks health metrics, and generates workout schedules that dynamically adapt to calf/shin pain signals using an AI coach.
Result: A live PWA-ready fitness platform featuring AI-generated personalized plans, real-time metric dashboards, and haptic feedback for a native-grade mobile experience.
Daily health metric capture with morning recovery score analysis to track and prevent injury escalation.
Real-time AI fitness advisor generating personalized workout schedules that adapt dynamically to pain signals.
Offline-capable, installable progressive web app with haptic feedback for a native mobile feel.
Problem: University timetabling is a highly constrained NP-hard problem — manual scheduling leads to clashes, faculty overloads, and wasted resources.
What I Built: An intelligent web-based timetable generation platform for higher education, submitted for Smart India Hackathon 2025, implementing constraint satisfaction algorithms to automatically resolve conflicts.
Result: A fully automated scheduler that handles faculty availability, room capacities, and course dependencies — generating conflict-free timetables in seconds.
Automatic detection and resolution of scheduling conflicts across faculty, rooms, and time slots.
Handles faculty availability, room capacities, lab requirements and consecutive period rules simultaneously.
Clean, printable timetable exports for departments and individual faculty members.
Problem: Computer Networks is one of the most concept-dense subjects in CS — students need fast, structured, exam-ready references that go beyond textbooks.
What I Built: A live interactive study companion web app with structured topic coverage, protocol breakdowns, visual diagrams, and TCP/IP stack walkthroughs — all in one deployable site.
Result: A live Vercel-deployed study tool that serves as a quick-reference handbook for CN topics including OSI model, TCP congestion control, DNS, HTTP, and subnetting.
Structured coverage of TCP, UDP, DNS, HTTP/HTTPS, SMTP, and more with visual protocol diagrams.
Interactive layer-by-layer walkthroughs with real packet flow examples and exam-ready summaries.
Subnetting cheatsheets, port number tables, and congestion control diagrams all in one place.