Dipendra Thapa — Software Engineer

Featured work

Selected projects

Real systems. Measurable outcomes. Production-ready.

⭐ Featured

🧠

CPU-Only Multi-Task AI System

End-to-end deep learning pipeline predicting age, gender, and ethnicity simultaneously. Eliminates GPU dependency — built for real-world production deployment with full reproducibility pipeline.

<100ms inference / prediction

3-in-1 tasks per forward pass

0 GPU dependency

Python TensorFlow Multi-task Learning CPU Inference Research

🧠

Brain Tumor Detection

MRI classification using MobileNetV2 with 89.3% accuracy. Features Grad-CAM explainability.

→ 89.3% acc · Grad-CAM XAI

TensorFlowMobileNetV2Streamlit

🎵

VibeStream

Distraction-free YouTube audio player. Real-time queue, sub-second loads, minimal UI.

→ <1s load · real-time queue · Vercel

ReactTypeScriptZustandVite

⚡

SolPayX

Decentralized remittance on Solana. Instant USDC transfers, near-zero fees, no banks.

→ Near-instant tx · ~$0.00025 fee

SolanaWeb3USDCDeFi

❤️

Research Paper

Heart Disease Predictor

ML classifier on WHO STEPS Nepal dataset. SMOTE-balanced, clinically relevant features.

→ Improved minority class recall via SMOTE

PythonStreamlitSMOTEML

More projects on GitHub

Research & Writing

Published Work

Peer-reviewed research and technical writing.

📄 Research Paper

CPU-Only Multi-Task Learning for Age, Gender & Ethnicity Prediction

TensorFlowMulti-taskCPU InferenceResearchGate

2025 Read paper

📄 Research Paper

Prediction of Heart Disease Using Machine Learning: WHO STEPS Nepal 2019 Dataset

PythonSMOTEMLResearchGate

2025 Read paper

About me

Engineer by craft,
curious by design.

I'm a Software Engineering student at Pokhara University focused on systems that actually work in production — not just on paper. My work spans CPU-optimized ML inference pipelines, real-time full-stack applications, and decentralized finance infrastructure.

I care about measurable performance: inference latency, transaction throughput, load time. If it can't be benchmarked, it's not done.

4+

Production-grade projects

<100ms

CPU-only AI inference

3

ML models in research

GitHub Activity — Last 90 days

Less

More

Experience

Journey so far

🎓

Dec 2022 — Present

B.E. Software Engineering

Pokhara University, Nepal — Focused on systems programming, machine learning, and distributed systems. CGPA: 3.7+

❤️

2025

Heart Disease Predictor — Research

Published ML research using WHO STEPS Nepal 2019 dataset. Applied SMOTE for class balancing. Deployed with Streamlit.

🧠

2025

CPU-Only Multi-Task AI — Published Research

End-to-end deep learning pipeline published on ResearchGate. Simultaneously predicts age, gender & ethnicity with <100ms CPU inference.

⚡

2026

SolPayX — Solana DeFi

Built decentralized remittance on Solana mainnet. USDC transfers with near-zero fees (~$0.00025) deployed to Vercel.

🎵

2026

VibeStream — Full-Stack Web App

YouTube audio player with real-time Firebase queue, YouTube Data API v3 search, and sub-second load times. Live on Vercel.

🏥

2026

Brain Tumor Detection — Deep Learning

Built an end-to-end MRI classifier using MobileNetV2. Achieved 89.3% accuracy and integrated Grad-CAM explainability in a Streamlit app.

Skills

Tech stack

Pokhara, Nepal

drag to rotate

AI / ML

Python TensorFlow Machine Learning SMOTE Multi-task Learning

Frontend

React TypeScript JavaScript Zustand Vite HTML / CSS

Systems / Other

C++ Flutter Firebase Git Solana / Web3

Testimonials

What people say

❝

"Dipendra demonstrates an exceptional ability to translate research concepts into production-ready systems. His work on CPU-only inference is genuinely novel."

👨‍🏫

Biswash Devkota

Software Engineering, Pokhara University

❝

"His attention to measurable performance — latency, throughput, load time — sets him apart. Every project ships with benchmarks."

🧑‍💻

Narayan Adhikari

Software Engineering, Pokhara University

❝

"From Solana smart contracts to ML pipelines, Dipendra builds across the full stack with a systems mindset that is rare at this stage."

🚀

Ansh Gurung

Gandaki Engineering College (GEC)

Contact

Let's build something serious.

Open to internships, research collaborations, and real-world impact projects. I respond fast.

GitHub LinkedIn

Building CPU-efficient AI that ships to production.

CPU-Only Multi-Task AI System

Let's build something serious.