Open to research & projects 10:42
mahesh_panta.cv
User Avatar
Computer Engineering ยท Kathmandu University

Mahesh Panta

Computer Engineering undergraduate at Kathmandu University with hands-on experience in machine learning, computer vision, and 3D reconstruction research.

๐Ÿ“
Summary

Computer Engineering undergraduate at Kathmandu University with hands-on experience in machine learning, computer vision, and 3D reconstruction research. Currently a Research Trainee at KU's HI Lab, working on depth estimation and 3D scene reconstruction. Proficient in building deep learning pipelines (TensorFlow, PyTorch), training object detection models (YOLOv8), and deploying ML-powered APIs with FastAPI and PostgreSQL. Passionate about applying computer vision, reinforcement learning, and full-stack AI to real-world problems.

๐Ÿงญ
Experience
Mar 2026 โ€”
Present
Research Trainee โ€” Computer Vision & 3D Reconstruction
HI Lab, Kathmandu University โ€” Dhulikhel, Nepal
  • Conducting research in computer vision and 3D scene reconstruction as part of KU's Human-Intelligence Lab.
  • Working on depth estimation, point cloud generation, and computer vision pipelines using Python and OpenCV.
  • Collaborating with lab members on applied CV research targeting real-world spatial understanding problems.
๐Ÿ—‚๏ธ
Projects
ChitraDartaa Civic Issue Reporting System GitHub โ†—
Flutter ยท FastAPI ยท PostgreSQL ยท YOLOv8 ยท TensorFlow
  • Built a full-stack civic issue reporting platform enabling citizens to submit geo-tagged image reports (potholes, garbage, road damage) with GPS location and resolution status tracking.
  • Trained YOLOv8 object detection and segmentation models to automatically classify issue types from uploaded images.
  • Developed CNN-based classifiers (road-issue vs non-issue, potholes vs garbage) with TensorFlow/Keras, applying data preprocessing and augmentation pipelines.
  • Designed a FastAPI REST backend with PostgreSQL (via SQLAlchemy) for persistent storage of reports, user data, and resolution tracking.
  • Built a cross-platform Flutter mobile frontend with interactive map integration, push notifications, and an admin dashboard for authorities.
Clothing Visual Search & Matching System GitHub โ†—
PyTorch ยท TensorFlow ยท FastAPI ยท React ยท PostgreSQL
  • Co-developed a full-stack visual search system that scans an uploaded clothing image and retrieves the closest matching item from a labeled wardrobe dataset spanning 25 categories.
  • Benchmarked ResNet, a ResNet encoder with CNN decoder, MobileNet, and YOLO architectures for feature extraction, training with metric learning (triplet loss, online batch-hard mining) to build a similarity embedding space.
  • Applied a Linear UCB (contextual bandit) strategy within the matching pipeline to guide result ranking.
  • Auto-generated YOLO-format labels with LangSAM and used class-balanced oversampling to improve accuracy on visually confusable categories.
  • Built a FastAPI + PostgreSQL backend with JWT authentication and a React frontend for uploading scans and browsing matches.
Forest Conservation System GitHub โ†—
Python ยท Computer Vision ยท Machine Learning
  • Developed an AI-powered forest monitoring system for detecting deforestation and environmental threats from image and remote sensing data.
  • Applied computer vision techniques for change detection and classification of forest cover using deep learning models.
  • Integrated ML pipelines for automated environmental risk assessment.
Game Vault: Mini Game Arcade with AI Agents GitHub โ†—
C++ ยท Raylib ยท ENet
  • Co-developed a multi-game arcade (Pong, Snake, Space Invaders) with real-time multiplayer via ENet reliable UDP networking.
  • Implemented RL-inspired AI agents: a Simple Reflex Agent for reactive play and a Model-Based Agent for stateful environment tracking.
  • Managed cross-platform builds and dependency resolution using CMake.
YummY-Chef Chatbot GitHub โ†—
Python ยท Kivy ยท ChatterBot ยท SQLite
  • Built a mobile-friendly chatbot providing recipe suggestions and cooking assistance using ChatterBot NLP and a Kivy cross-platform UI.
  • Implemented persistent data storage and conversation memory with SQLite.
๐Ÿ†
Achievements
๐Ÿ†
Locus HACK-A-WEEK Finalist (Pulchowk LOCUS Hackathon 2026)
LinkedIn โ†—
Advanced to the finals among top competing teams with the Forest Conservation System project.
โš™๏ธ
Skills
Programming Languages
PythonCC++
ML & AI
Supervised & Unsupervised LearningCNNsTransfer Learning Data AugmentationModel OptimizationRecommender Systems
Computer Vision
YOLOv8OpenCV3D Reconstruction Depth EstimationStructure-from-MotionNeRF MobileNetMetric Learning
Reinforcement Learning
Q-LearningModel-Based AgentsReflex Agents Environment SimulationContextual Bandits (UCB)
Deep Learning Frameworks
TensorFlowKerasPyTorch (fundamentals)
Data Science
NumPyPandasMatplotlibScikit-learn
Backend & APIs
FastAPIFlaskREST API Design
Databases
PostgreSQLSQLiteSQLAlchemy
Mobile & Frontend
Flutter (Dart)KivyReact HTMLJavaScript (Beginner)
Game Development
C++ with RaylibRayGUI
Networking
ENet (Reliable UDP)REST APIs
Tools & OS
GitCMakeUbuntuJupyter Notebook
Languages
EnglishNepaliHindiJapanese
๐ŸŽ“
Education
Kathmandu University
Bachelor of Engineering in Computer Engineering โ€” Dhulikhel, Nepal
Present (3rd Year)
Khwopa Secondary School โ€” Nepal
NEB Plus Two (Science with Biology) | GPA: 3.7
Graduated
๐Ÿ“œ
Certifications
Unsupervised Learning, Recommenders, Reinforcement Learning
Deep Learning.AI
Apr 2026
49W64MVQ5LQ9
Machine Learning Specialization
Deep Learning.AI
Apr 2026
JUBLWYJDIHKT
Advanced Learning Algorithms
Deep Learning.AI
Nov 2025
BQCH8WWP8VMT
Supervised Machine Learning: Regression and Classification
Deep Learning.AI
Oct 2025
U0574LHEIT2T
Introduction to Programming Using Python
CS50, Harvard University
โ€”
Pandas
Kaggle
โ€”
Save CV
โ—†
Summary Experience Projects Achievements Skills Education Certs