Samir
Wagle
Wagle
About
Hey There!
I am an AI and NLP Engineer specializing in deep learning, multilingual language technologies, and data-driven problem solving. My work spans research, model development, dataset creation, and system-level engineering ranging from low-resource NLP to embedded device integration and Android-based POS systems.
Beyond AI research and software engineering, I explore how emerging technologies, automation, and data-driven intelligence can shape industries from finance and business to IoT and public platforms. With an engineering mindset and a strong foundation in applied machine learning, I aim to bridge the gap between academic research and practical deployment.
If you are interested in AI, NLP, system engineering, POS devices or industry-focused innovations, feel free to connect. I am always excited to collaborate, share ideas, and build meaningful solutions.
Portfolio
Selected Works
NepSAUL — Nepali Legal QA System
🏆 1st Place, aIDEA National AI Hackathon 2026 (NPR 100,000). RAG pipeline over 10,000+ Nepali Supreme Court case laws using BM25 + multilingual-e5-large + FAISS, achieving 91% Precision@1. Evaluated with LLM-as-Judge; migrating to an agentic n8n framework for multi-step legal reasoning.
NepMed — Mental Health Compass
An innovative mental health and medical platform developed during the Nepal-US Hackathon 2026. The system aims to provide navigational support and mental wellness assessments with personalized guidance.
Nepal Election Analysis
ML-driven analysis of Nepal's 2022 federal and provincial election data, combining web-scraped results with NLP-based voter sentiment analysis from social media. Built an interactive dashboard visualizing party-wise seat distribution, constituency-level vote shares, and regional political trends. Applied clustering and classification models to identify swing constituencies and voter behavior patterns across all 7 provinces. Delivers data-driven insights into Nepal's evolving political landscape through reproducible Python pipelines.
Nepali Profanity & Offensiveness Detection
Production NLP system detecting profanity, offensiveness, and speaker gender in Nepali text. Bi-LSTM + multilingual BERT on 11.6K+ annotated comments, 87.8% accuracy. FastAPI backend with rate limiting, Next.js + Three.js frontend. Published at ICON 2024.
NEPSE All Scraper — Nepal Stock Exchange Pipeline
Open-source automated data pipeline that scrapes stock prices, dividends, right shares, and floorsheet data for all 337 companies listed on the Nepal Stock Exchange (NEPSE). Runs on a scheduled GitHub Actions workflow, delivering clean, versioned financial datasets without manual intervention.
NECPrep — NEC License Exam Platform
Free, open-source exam preparation platform for Nepal Engineering Council (NEC) license candidates. Features 1,380+ questions across 10 topics, timed mock tests, flashcards, bookmarks, progress analytics, and an immersive 3D landing page — all fully client-side with no account required.
Explore more projects spanning AI/ML, NLP, data engineering, embedded systems, and web development.
Research & Publications
Research Work
MEME-Fusion@CHiPSAL 2026: Multimodal Ablation Study of Hate Detection and Sentiment Analysis on Nepali Memes
We propose a hybrid cross-modal attention fusion system for the CHiPSAL 2026 shared task on hate speech detection (Subtask A) and sentiment classification (Subtask B) in Nepali memes. The architecture combines CLIP (ViT-B/32) for visual encoding with BGE-M3 for multilingual text representation, fused through 4-head self-attention and a learnable gating network that dynamically weights modality contributions per sample. Evaluation across eight model configurations demonstrates explicit cross-modal reasoning achieves a 5.9% F1-macro improvement over text-only baselines, while revealing that English-centric vision models exhibit near-random performance on Devanagari script and that ensemble methods catastrophically degrade under low-resource data scarcity.
Retrieval-Augmented Generation Framework for the Nepali Legal Domain Question Answering
The first application of a Retrieval-Augmented Generation (RAG) pipeline for Nepali legal question answering, using 10,320 Supreme Court case laws from the NKP digital archive. Comparing sparse (BM25) and dense (multilingual-e5-large, e5-base, LaBSE) retrieval strategies, BM25_Chunks achieved the highest Precision@1 of 91%. Generation using BM25_Docs yielded 85% truthfulness, 74% groundedness, and a 92% answer generation rate, establishing the first baseline RAG framework for low-resource Nepali legal NLP.
Evaluating Sentence Embedding Models for Nepali Sentiment Analysis: A Comparative Study
A rigorous comparative analysis benchmarking four state-of-the-art multilingual sentence embeddings — BGE-M3, LaBSE, mE5-base, and DistilUSE — across three neural classifiers (MLP, Residual MLP, Transformer) for Nepali sentiment analysis. BGE-M3 paired with a simple MLP achieved 82.49% accuracy, demonstrating that embedding quality is the dominant performance determinant over architectural complexity in low-resource NLP settings.
Profanity and Offensiveness Detection in Nepali Language Using Bi-Directional LSTM Models
We present a Bi-directional LSTM-based approach for detecting profanity and offensive content in Nepali social media text. Using a custom annotated dataset scraped from YouTube, the model achieves strong classification performance on low-resource Nepali, demonstrating the effectiveness of BiLSTM architectures for colloquial and code-mixed Nepali NLP tasks.
Services
My Services and Field
Data Driven Research and Development
I specialize in cutting-edge research and development, leveraging Python and machine learning to create innovative solutions. My expertise spans natural language processing (NLP), where I build advanced models for text analysis, sentiment detection, and automation. I have worked on profanity detection systems, ensuring safer digital communication through AI-driven content moderation. I design and evaluate Retrieval-Augmented Generation (RAG) pipelines — combining sparse and dense retrieval with large language models — for domain-specific question answering in data-scarce, low-resource settings. With a strong foundation in research, I continuously explore new methodologies to enhance AI-driven solutions for diverse applications.
AI Chatbots & RAG Pipelines
Expert in building production-grade AI chatbots and Retrieval-Augmented Generation (RAG) systems using Azure OpenAI and OpenAI APIs. I architect end-to-end pipelines — from document ingestion, chunking, and vector indexing to context-aware multi-turn conversation flows. Experienced in integrating LLMs into real-world applications with custom system prompts, grounding, and guardrails to ensure accurate, trustworthy responses.
Workflow Automation & Cloud Infrastructure
Proficient in building intelligent no-code and low-code automation workflows using n8n, integrating AI agents, webhooks, APIs, and databases into seamless pipelines. Experienced with Microsoft Azure — deploying and managing Azure VMs, configuring Azure OpenAI services, and working within the Azure portal for resource management and access control. Also hands-on with Google Cloud Console for cloud-hosted model serving and data pipelines.
Web Development
Experienced in full-stack web development, specializing in building scalable, user-friendly applications. Proficient in front-end and back-end technologies, including JavaScript, React, Node.js, and Python. Passionate about creating responsive, high-performance websites and APIs, integrating modern frameworks, and optimizing user experiences.
Designing
Experienced in UI/UX design, creating intuitive and visually appealing layouts using Figma, Canva, Photoshop, and Lightroom. Skilled in crafting responsive and user-friendly designs for web and mobile applications.
Robotics, IoT and Embeded System
Passionate about developing intelligent automation systems by integrating robotics with IoT. Skilled in sensor integration, embedded systems, and real-time data processing to create smart, connected solutions for various applications.
Photography and Videography
Experienced in capturing high-quality visuals with a keen eye for composition, lighting, and storytelling. Proficient in photo and video editing using Photoshop, Lightroom, and Premiere Pro to create engaging and professional content.
Social Media Handelling
Mastering the art of social media, I navigate the digital landscape, harnessing its potential to drive engagement, cultivate brands, and create meaningful connections with target audiences. A youtube SEO and Analytics Expert for your growth and increase in reach.
Experiences
Career Highlights
Freelancer & Remote AI Consultant
● Built and deployed custom AI chatbots using Azure OpenAI and OpenAI APIs, tailored for client-specific use cases including customer support, document Q&A, and internal knowledge bases.
● Designed and implemented end-to-end automation workflows using n8n, integrating AI agents, webhooks, REST APIs, and databases to automate repetitive business processes.
● Provisioned and managed Azure Virtual Machines for hosting AI services and backend APIs, configuring networking, security groups, and deployment pipelines.
● Worked with Azure OpenAI Studio and Google Cloud Console for model deployment, endpoint configuration, and cloud resource management.
● Delivered RAG-based solutions combining document retrieval with LLM generation for intelligent, grounded responses over client datasets.
Research And Development Engineer (Software Engineering )
● Worked on embedded C programming for the POS payment terminal powered by the Asino Q161 Pro microcontroller.
●Designed and implemented user interface features including image rendering, audio playback, and interactive
screens for the POS terminal.
●Developed and tested TCP socket communication modules for real-time data exchange, including sending/receiving
JSON data to/from backend servers.
● Worked on Android Based POS Device, Migrated the entire code base from Android 10 to Android 14
resolving compatibility issues, updating libraries,SDK’s and system level components and ensuring full
compatibility with new OS environment
● Diagnosed and fixed major hardware integration issues including card reader, NFC, and PIN pad failures
by collaborating with the manufacturer, updating SDKs, and modifying device-level code
● Performed extensive bug fixing, optimization, and system verification to ensure a stable, production-ready
POS workflow.
● Redesigned the UI for a 240×320 px POS display, improving usability on a small form-factor device
Research Intern (Undergraduate)
● Developed deep learning models for sentiment classification and created an English Text
dataset by scraping
YouTube comments for Sentiment Analysis
● Published a Research Paper titled ” Profanity and Offensiveness Detection in Nepali
Social Media Using
Bi-directional LSTM Models ” at 21st International Conference of Natural Language
Processing( ICON
2024), MIT Campus of Anna University
● Contributing to the development of Trilingual Machine Translation System (
English-Nepali-Tamang) in
Society-Centered AI Research Program by Google
Co Founder / Technology and Innovation Co Ordinator
● Administrate and manage the Microsoft Intra Admin Center, ensuring secure and efficient
access manage-
ment, identity governance, and compliance
● Oversee user access controls, authentication policies, and security protocols while
providing technical support
and streamlining IT operations to enhance infrastructure and improve organizational
efficiency.
● Conduct extensive online research to identify and apply for nonprofit services, grants,
and support pro-
grams by analyzing eligibility criteria, preparing the necessary documentation, and
streamlining application
processes to maximize opportunities for nonprofit growth and sustainability.
● Conduct online research to identify nonprofit services, grants, and funding
opportunities, while analyzing
eligibility criteria, preparing documentation, and writing compelling proposals to secure
financial support
and streamline application processes for nonprofit initiatives
● Handled Fiscal Partner Workload for UbuCon ASIA 2025.
Public Relation officer and Social Media Manager
● Managing Club's Public Image, Relation and Digital Presence.
● Designing of Logos, Posters, Flexes and Social Media Post.
● Website and E-Mail Handelling
Education
Academic Credentials
Bachelor's in Engineering - Computer Engineering
+2/ Intermediate
Primary and Secondary Schooling ( SEE)
License and Certifications ( Please visit LinkedIn for Credentials and Verifications )
Verified Credentials.
Understanding Machine Learning
Data Manipulation with Pandas
Full Stack Web Development with Flask
C: The Mother Language
C++ For the Rest of Us
Introduction to Python
Skills
Areas of Expertise
Python
C
C++
Java
JavaScript
Kotlin
TensorFlow
PyTorch
Hugging Face
Android
React
Next.js
Node.js
Flask
CSS3
Git
Figma
Photoshop
Illustrator
Canva
VS Code
GitHub
From Friends and Co- Workers
Testimonials
Contact
Let’s Talk
Open to research collaborations, freelance AI/NLP projects, and interesting conversations. Drop a message — I respond quickly.
- Satdobato, Lalitpur, Nepal
- [email protected]
- +977 984-003-2620