Hi, I'm Bhupendra Prasad Sah
Computer Engineering Student — Full-Stack & AI Engineer
I'm a Computer Engineering student passionate about Full-stack development, Data science, and AI / Machine Learning & Deep Learning. I enjoy combining software engineering with AI to build practical solutions to real-world problems.
IOE Purwanchal Campus
B.E. Computer Engineering
Featured Engineering Projects
Real-world systems with quantifiable results. Explore problem formulations, machine learning methodologies, and frontend performance metrics.
Intelligent Multi-Agent Tourism Platform (TravelNepal)
Agentic Itinerary Planner with 5-Tier RBAC, Emergency SOS & Fintech Integrations
The Problem
Travelers in Nepal frequently face disconnected booking experiences, static unverified estimates, lack of multi-role coordination across vendors, and hazardous isolation during remote or high-altitude emergencies.
Engineering Solution
Engineered an autonomous multi-agent travel platform using FastAPI, Google GenAI SDK (Gemini), and LangGraph querying verified PostgreSQL records for 100% bookable itineraries, coupled with Next.js 16 Server Actions, Khalti/eSewa payment gateways, and a 1-Click SOS rescue protocol.
Quantifiable Impact & Metrics
- 100% bookable, budget-aware multi-day travel plans verified against real database inventory
- 5-tier RBAC security across tourists, hotel owners, restaurants, guides, and admins
- 1-Click instant emergency SOS protocol broadcasting real-time GPS coordinates
System Architecture & Implementation Details:
- •AI Multi-Agent Planner: Built an agentic itinerary system using FastAPI, Google GenAI SDK (Gemini), and LangGraph querying verified PostgreSQL records for 100% bookable, budget-aware multi-day travel plans.
- •Full-Stack Architecture: Engineered a Next.js 16 frontend with Server Actions, Tailwind CSS, and 5-tier Role-Based Access Control (RBAC) across tourists, hotel owners, restaurants, guides, and admins.
- •Commerce & Safety: Integrated Khalti & eSewa payment gateways, in-trip expense tracking, and a 1-Click SOS instant emergency protocol dispatching live geolocation to rescue authorities.
- •Database Layer: PostgreSQL with Drizzle ORM and SQLAlchemy 2.0 ensuring strict transactional integrity across bookings and payments.
Agri-XAI – Multimodal Crop Diagnostic & Agronomic RAG System
Deep Learning Diagnostic Pipeline with Visual Attention Heatmaps & Multimodal Advisory
The Problem
Crop yield loss threatens agricultural viability due to delayed disease identification, while generic generative AI models frequently hallucinate incorrect or hazardous chemical pesticide dosages.
Engineering Solution
Constructed a high-precision diagnostic and advisory system using PyTorch (EfficientNet-B4) with Grad-CAM backprop hooks for visual explainability, paired with Neon PostgreSQL pgvector (HNSW cosine indexing) and Gemini 1.5 Flash for verified, zero-hallucination agronomic guidance.
Quantifiable Impact & Metrics
- Zero-hallucination vector retrieval validated against verified FAO/ICAR agricultural database standards
- Automated foliar lesion pinpointing and precise leaf infection percentage calculation
- Multilingual audio speech synthesis (TTS) paired with actionable active ingredient prescription
Vingo – Multi-Role Food Delivery Platform
Multi-Stakeholder Ordering Platform with Dynamic Redux Cart & Cloud Database
The Problem
Modern delivery operations demand high synchronization across customers, restaurants, couriers, and administrators, where poor state consistency causes stale orders and inventory clashes.
Engineering Solution
Engineered an end-to-end multi-stakeholder food delivery web application interconnecting Customers, Restaurant Owners, Delivery Partners, and Admins with protected routes, dynamic Redux state, and secure JWT/OAuth authentication.
Quantifiable Impact & Metrics
- 4 distinct synchronized user roles: Customers, Restaurant Owners, Delivery Partners, and Admins
- Dynamic client cart and persistent checkout flow driven by Redux toolkit
- Cloudinary-backed media pipeline and modular REST endpoints deployed on Vercel
Hospital Readmission Risk Prediction
Clinical Decision Support & Machine Learning Diagnostic Pipeline
The Problem
Unplanned hospital readmissions cost healthcare systems billions of dollars annually and jeopardize patient health. Early risk detection enables clinical teams to implement targeted discharge interventions.
Engineering Solution
Engineered an end-to-end Machine Learning pipeline utilizing scikit-learn, advanced feature engineering, class imbalance mitigation, and cross-validation to predict readmission likelihood with high precision.
Quantifiable Impact & Metrics
- 85% predictive accuracy on out-of-sample patient validation data
- 20% reduction in false positives through precision-recall threshold calibration
- High clinical interpretability through feature importance attribution
Resume, Skills & Qualifications
Directly mapped to Bhupendra's verified resume: technical stack, academic trajectory at IOE Purwanchal Campus, professional certifications, and recruiter Q&A.
Programming Languages
Solid foundations in systems programming, strongly-typed scripting, and database query languages.
Full-Stack & Web
Modern, reactive, accessible, and high-performance server/client web architectures.
AI, ML & Computer Vision
Deep learning backbones, explainable attention hooks, agentic workflows, and LLMs.
Databases & Vector Search
Serverless relational data stores, high-dimensional vector indexes, and modern ORMs.
DevOps & Security
Containerization, version control, role-based authorization, and fintech payment gateways.
Need an offline copy of Bhupendra's updated resume?
Download the single-page ATS-formatted PDF resume updated with TravelNepal, Agri-XAI, and Vingo.
Let's Build Something Extraordinary
Whether you have an engineering opening, a freelance project, or an innovative AI venture, my inbox is always open.
Primary Email
Direct inquiries & interview invitations