Pranish Kafle

B.E. in Computer Engineering

Kathmandu University · 2021–2025

Muzzle-Based Livestock Identification via Siamese Deep Metric Learning

2024 – present

• Contributed to a **Plantsat** initiative in collaboration with the **United Nations Development Programme (UNDP) in Nepal**, aimed at scalable livestock identification for farmers and agricultural programs.
• Designed and compared Siamese Network architectures with **VGG16, ResNet50, ResNet152, ViT-Base, and ViT-Large** backbones; twin branches share weights to generate high-dimensional image embeddings for identity verification.
• Conducted multi-environment field testing across multiple farms in Nepal, discovering that **single-condition muzzle datasets are insufficient** — the model required images from diverse lighting, angles, and environmental contexts to generalize well.
• Collaborated with a slaughterhouse to collect muzzle images at controlled post-mortem intervals (**12 hr before, 2 hr before, immediately before, immediately after, and 2 hr after slaughter**), capturing how muzzle shape compresses and changes over time, enabling the model to adapt to deformation conditions.
• Built and deployed a dedicated data-collection application to capture muzzle images across all farm sites and time intervals, enabling **structured large-scale annotation**.
• Optimized embedding representations in a **128-dimensional latent space** via **Contrastive Loss (margin = 2.0)**, minimizing distance for same-identity pairs and pushing different-identity pairs beyond threshold.
• Integrated **Gradient-weighted Class Activation Mapping (Grad-CAM)** to visually confirm that network attention focuses on anatomical muzzle groove patterns rather than background noise — a critical step for **model trustworthiness and academic validity**.
• Evaluated binary classification performance using **accuracy, precision, recall, F1-score, and confusion matrices** by tuning Euclidean distance decision boundaries.

An Empirical Comparison of Embedding Models for Nepali Legal Document Retrieval

Abhiyan Dhakal, Kausik Paudel, Pranish Kafle, Sugat Sujakhu, Prakash Poudyal, Anita Jadhari

ICAIL 2026 · 2026 · Short paper – presented 9 June 2026

Compares retrieval behavior rather than treating embedding choice as an implementation detail, with attention to the constraints of Nepali legal text.

Official programme

Detecting Image Forgeries and Deepfakes: A Comparative Study of CNN and Transformer Models with a Custom-Curated Dataset

Pranish Kafle

NCCI 2025, Dhulikhel, Nepal · 2025 · Published

Created a custom-curated dataset and evaluated CNN versus Transformer architecture performance on image forgery and deepfake detection.

Feasibility of Artificial Intelligence Driven Analysis in the Context of Nepalese Legal System

Abhiyan Dhakal, Sugat Sujakhu, Pranish Kafle, Kausik Paudel, Prakash Poudyal

Proceedings of ICAIL 2025, pp. 498–499 · 2025 · Published

Processes legal documents for retrieval, expands user queries, and evaluates rule-recall, rhetorical-understanding, and interpretation-oriented questions.

DOI

Software Engineer (ML & Computer Vision)

Plantsat · Aug 2024 – Present

• Led the design and evaluation of **Siamese neural network architectures** (VGG16, ResNet50, ResNet152, ViT-Base, and ViT-Large) utilizing **Contrastive Loss** for non-invasive biometric livestock identification.
• Integrated **Explainable AI (Grad-CAM)** to visually map and validate that neural activation focus was anchored on anatomical muzzle groove ridges instead of background noise.
• Engineered **on-device image processing pipelines** for real-time camera frames, including auto-cropping, blur-filtering, and illumination normalization.
• Optimized deep learning model inference (onnx runtime / tflite integration) for resource-constrained edge devices, achieving low-latency offline verification.
• Coordinated large-scale **multi-environment field data collection** across Nepalese farms, capturing muzzle characteristics under variant ambient lighting, postures, and post-mortem intervals.

Junior Software Engineer

Emberce Technology · Feb 2023 – Aug 2024

• Developed and integrated on-device image analysis components, including **computer vision modules for localized skin feature classification**.
• Optimized client-side networking layers, caching strategies, and local database transactions to ensure efficient retrieval of image metadata.
• Designed responsive interface architectures to handle high-resolution image rendering pipelines across various hardware form factors.

Medease — Healthcare Management & Booking System

An integrated healthcare management application featuring real-time appointment booking, consultation chat, video consulting, and a built-in medical products e-commerce system.

Stack: Flutter · Kotlin · WebRTC · Firebase · Node.js

Lunar Calibration for Light Pollution Measurement

A collaborative Android application developed with students from MIT and NYU that detects and measures light pollution from lunar photographs. The app handles camera-based image capture, analysis, and all computation entirely on-device.

Stack: Android · Kotlin · OpenCV · On-Device ML

Mobile App Development Community Coordinator

Kathmandu University Computer Club · 2024–2025

Coordinated mobile app development initiatives and workshops at Kathmandu University.

Mentor

KU Hackfest · 2024–2025

Mentored participants at the biggest in-person hackathon in Nepal.

Winner

Hack For Business Hackathon · 2024

Won the Hack For Business Hackathon.