Projects

Backend Developer & Engineer

Hetauda School of Management and Social Sciences had no unified digital presence - academic content, admissions, events, clubs, and notices were scattered across disconnected channels with no central management interface, no audit trail, and no way for staff to update content without developer involvement.

As Backend Developer & Engineer, designed and implemented the server-side architecture using Next.js 16 App Router with a clean layered architecture (Schema → Repository → Service → DTO → API Route → Page). Owned the backend systems including the versioned REST API, database schema, middleware layer, and RAG pipeline integration alongside a role-gated admin dashboard for content management.

Covers the institution's complete digital footprint: public website with 15+ content modules, admin CMS with CRUD for all content, a centralized MinIO media library, versioned REST API with Swagger UI, IP-based rate limiting, and a RAG-powered AI chatbot - all deployed at https://hsm.prabeshguragai.com.np/

Lead Full-Stack Developer

Schools managing student records, attendance, portfolios, and exam results across multiple programs and batches face fragmented tooling - spreadsheets, paper registers, and disconnected systems make it impossible to enforce workflows, maintain audit trails, or give students and teachers a unified view of academic progress.

Built a full-stack academic management platform for Hetauda School of Management & Social Sciences using Django REST Framework and React 19 + Tauri 2. Implemented a service-selector-serializer backend architecture across 8 Django apps covering students, academics, attendance, portfolios, and audit logs - deployed as both a web app and a native cross-platform desktop application.

Handles end-to-end institutional operations: auto-generated student IDs, hierarchical academic structures (Program → Batch → Semester → Section → Subject), portfolio submission approval workflows, QR-coded ID cards, and a full audit trail with JSON diffs

Solo Project

SaaS platforms need strict data isolation between tenants, a single misconfigured query can expose one client's data to another, making multi-tenancy one of the hardest correctness problems in backend engineering.

Built a production-ready blogging API with tenant-scoped data isolation at the query layer, JWT & Google OAuth2 authentication, a real-time notification delivery system, and an AI-powered content recommendation engine.

90%+ test coverage verified by Pytest. Redis caching eliminates redundant DB hits on hot endpoints. Deployed on Render with zero cross-tenant data exposure confirmed across the full test suite.

Solo Project

REST polling for messaging creates latency, wasted requests, and a broken UX teams need true real-time delivery with typing indicators, read receipts, and proper group permission hierarchies.

Engineered a full-stack chat platform using Django Channels for persistent WebSocket connections, with Redis as the channel layer for horizontal scaling. Implemented private and group conversations, owner/admin/member RBAC, and full-text message search.

WebSocket connections replace polling entirely messages are pushed instantly with no client-side request overhead. Role-based permissions enforced at the channel layer prevent privilege escalation.

Solo Project

Shared docs tools like Google Docs offer no tenant isolation, no audit trail, and no fine-grained access control teams can't know who changed what, or enforce meaningful permission boundaries.

Built a multi-tenant documentation platform with organization-scoped workspaces, RBAC (owner/admin/member/viewer), document versioning with immutable audit logs, invite-based team management, and JWT access + refresh token auth.

Each organization is fully isolated no cross-tenant data leakage by design. Every write is captured in an immutable audit log, making the platform compliance-ready out of the box.

Solo Project

Streaming apps transmit full video data even when you only want audio burning bandwidth on pixels you never see. Browser-based players can't offer a native desktop experience or work well on low-bandwidth connections.

Created a minimalist cross-platform desktop music streaming app that uses yt-dlp for audio-only extraction, cutting out video data entirely. Features trending music discovery, keyword search with tag filtering, and a FastAPI backend for API routing and audio processing.

Audio-only streaming slashes bandwidth consumption dramatically compared to full video streams a meaningful difference on metered or slow connections. Ships as a native desktop app on Linux, Windows, and macOS with no browser dependency.