Download Resume
Open to Roles — Available Now

Sahil Duwal

Computer Engineer AI / ML Engineer Web Developer

Building intelligent systems at the intersection of AI, computer vision, and full-stack engineering. Turning rigorous research into impactful, production-grade software.

View Projects Download CV
Sahil Duwal
AI / ML
Computer Vision
8+ AI Projects
10+ Certifications
3+ Years Learning
Scroll
About

Computer Engineer
& AI Researcher

Computer Engineer specializing in deep learning architectures, computer vision, and full-stack web systems. Experienced building end-to-end AI pipelines — from dataset curation to model deployment. Passionate about engineering rigorous, real-world solutions at the frontier of AI research.

Currently pursuing AI-driven internships, research collaborations, and full-time roles where I can apply technical depth with product-level thinking.

See My Work
◆

Focus Areas

Python, TypeScript, JavaScript, AI/ML, Computer Vision, Full-Stack Web Development

■

Currently

Full Stack Development Trainee at CVArcher

○

Location

Bhaktapur, Nepal

▲

Open To

Internships, Research Collaborations, Full-time Roles

Expertise

Technical Skills

Five domains covering the full ML engineering and web development stack — from low-level systems to production deployment.

01

Programming Languages

Systems-level to scripting: the languages powering every project from day one.

Python TypeScript JavaScript C C++ HTML5 CSS3
02

AI, Machine Learning & Vision

End-to-end ML: architecture design, training pipelines, and model evaluation.

PyTorch Computer Vision Deep Learning Natural Language Processing (NLP) scikit-learn
03

Full-Stack Web Development

From REST APIs to reactive frontends — scalable, type-safe, production-ready.

Node.js React Next.js FastAPI Tailwind CSS RESTful APIs
04

Data Science & Databases

Statistical analysis, feature engineering, and relational data management.

Pandas NumPy Matplotlib MySQL Data Analysis Statistical Modeling
05

Developer Tools & Frameworks

Version control, CI workflows, and cross-platform dev environments.

Git GitHub Linux/Unix LaTeX Git Workflow
Experience

Professional Journey

May 2026 – Present Full Stack Development Trainee

Full Stack Development Trainee

CVArcher

  • Built and shipped responsive TypeScript/Node.js web applications with robust server-side logic and type-safe APIs.
  • Implemented Google OAuth authentication and third-party payment gateway integration, securing user sessions and enabling transactional flows.
  • Executed systematic QA and functional testing cycles, resolving critical bugs and improving application stability under real-world usage conditions.
Oct 2025 – Dec 2025 Data Annotation Intern

Data Annotation Intern

Next Wave AI

  • Annotated high-quality image datasets (objects and media covers) and generated structured JSON schemas for AMS360 to support computer vision model training and workflow automation.
  • Evaluated Large Language Model (LLM) security behaviors through targeted prompt testing and adversarial analysis, identifying key vulnerability patterns.
  • Researched and benchmarked emerging AI video generation tools, producing structured capability assessments for internal review.
2025 – 2026 DataCamp Fellow

DataCamp Fellow

Code for Nepal

  • Competitively selected for an intensive data science fellowship, accelerating expertise in Python, exploratory data analysis (EDA), and practical machine learning pipelines.
2024 – 2025 Tech Support Member

Tech Support Member – Hult Prize On-Campus Program

Khwopa College of Engineering

  • Delivered comprehensive technical support for event management, complex presentation setups, and digital logistics.
  • Assisted participating teams in troubleshooting IT-related and software issues during high-stakes workshops, mentoring sessions, and hackathons.
  • Ensured flawless technical operations across multiple campus-wide events, directly contributing to the program's operational success.
Laboratory

Research & Development

Rigorous AI/ML engineering projects with documented architectures, datasets, and evaluation metrics.

AI/ML · Flagship
↗

Deepfake Detection & Localization

Developed a hybrid ViGNN–U-Net model for detecting and localizing deepfake content with high accuracy and robustness across various manipulation types.

Model Architecture
Hybrid ViGNN–U-Net
Dataset
Multi-source face manipulation dataset
Key Metric
85% Detection Accuracy
💡 Key Insight

Hybrid architecture combining ViGNN's frequency analysis with U-Net's spatial localization achieved superior performance over single-branch models.

PyTorch U-Net ViGNN Computer Vision Deep Learning
AI/ML · NLP
↗

Quantum-Enhanced BERT Model

Hybrid classical-quantum ML pipeline combining multilingual BERT embeddings with variational quantum circuits (VQCs) for low-resource Nepali offensive-speech detection.

Model Architecture
BERT + Variational Quantum Circuits
Dataset
Nepali offensive speech dataset
Key Metric
Improved F1-score
💡 Key Insight

Quantum-enhanced embeddings capture richer semantic representations for low-resource languages.

PyTorch BERT Variational Quantum Circuits NLP
AI/ML · AVSR
↗

Shruti: AVSR in Nepali Language

Audio-visual speech recognition (AVSR) system built for the Nepali language, combining acoustic and visual cues for improved recognition accuracy.

Model Architecture
CNN + LSTM Audio-Visual Fusion
Dataset
Nepali AVSR dataset
Key Metric
85% Accuracy
💡 Key Insight

Combining audio and visual modalities significantly improves speech recognition in low-resource settings.

Deep Learning Computer Vision NLP Python
AI/ML · Chatbot
↗

Ghumti: AI Travel Assistant

Built an intelligent chatbot powered by Llama 3 to provide real-time local bus routes, schedules, and optimized travel information.

Model Architecture
Llama 3 + RAG
Dataset
Local transit APIs
Key Metric
Response time <2s
💡 Key Insight

Llama 3 enables efficient reasoning over structured transit data for real-time assistance.

Llama 3 NLP Chatbot API Integration
AI/ML · RAG
↗

PDF Q&A RAG Bot

Local retrieval-augmented generation (RAG) application enabling semantic Q&A across multiple PDFs, running entirely offline.

Model Architecture
Llama2 + LangChain + Embeddings
Dataset
User-supplied PDFs
Key Metric
<500ms Response Time
💡 Key Insight

Semantic chunking + RAG pattern eliminated hallucinations while maintaining sub-500ms latency for local deployment.

Llama2 LangChain DocArray Sentence-Transformers
AI/ML · Captioning
↗

Image Captioning Model

Image captioning model combining a pretrained ResNet50 encoder with an LSTM decoder to generate descriptive captions for input images.

Model Architecture
ResNet50 Encoder → LSTM Decoder
Dataset
Flickr8k (8K images + 40K captions)
Key Metric
BLEU-4: 0.32
💡 Key Insight

Beam search with length penalty prevented caption truncation; attention mechanisms localized salient image regions.

PyTorch ResNet50 LSTM Flickr8k
Web · Expense Tracker
↗

CentraSpend – Expense Tracker

Full-stack personal expense tracking application with account management and transaction history.

Tech Stack
React, Express, Node.js, MongoDB
Features
Account management, transaction history
Key Metric
~2sec Load Time
💡 Key Insight

Architected a MERN-stack personal finance tracker with RESTful transaction API, demonstrating end-to-end full-stack ownership with account management and multi-entry transaction history.

React Express Node.js MongoDB
Web · Recipe Portal
↗

RecipeBook: Cooking Recipe Portal

Web application for storing, organizing, and managing cooking recipes with a full client-server architecture.

Tech Stack
React, FastAPI, MySQL
Features
Recipe management, search, user authentication
Key Metric
~3sec Load Time
💡 Key Insight

Decoupled React SPA and FastAPI service layer demonstrate clean separation of concerns, enabling independent scaling of frontend and backend.

React FastAPI MySQL
Web · Portfolio
↗

Personal Portfolio Website — Portfolio 2.0

Personal portfolio website showcasing skills, experience, and completed projects; rebuilt as a v2 with an updated design.

Tech Stack
TypeScript, React, CSS
Features
Responsive design, project showcase, blog section
Key Metric
Load time <1.5s
💡 Key Insight

Engineered a custom GSAP animation system, directional scroll-reveal, and live GitHub API integration — achieving sub-1.5s load times with zero external UI frameworks.

TypeScript React CSS
Research

Technical Deep Dives

In-depth explorations of engineering challenges and solutions.

Optimizing Wav2Vec2 for Low-Resource Languages

2025 Intermediate

Fine-tuning speech recognition on limited Nepali datasets while maintaining acoustic robustness across dialects.

Challenge: Low-resource languages lack labeled audio. Direct fine-tuning overfits and fails across dialects.

Solution: Three-stage approach—self-supervised pretraining on unlabeled audio, supervised fine-tuning with SpecAugment, and accent-aware reweighting using SMOTE.

$L = \alpha L_{ctc} + (1-\alpha) L_{accent}$ where $\alpha=0.5$

Result: 18% WER (vs 32% baseline), 76% generalization across 5 dialects.

Wav2Vec2 Speech Augmentation

Handling Class Imbalance in Medical Image Segmentation

2025 Intermediate

Training U-Net on imbalanced datasets where tumor pixels comprise <1% of total pixels.

Challenge: Medical images have severe class imbalance; standard BCE loss biases toward background.

Solution: Focal Loss ($\gamma=2$) + Weighted Dice + patch-based sampling + post-hoc thresholding.

$FL(p) = -\alpha_t(1-p_t)^{\gamma} \log(p_t)$

Results: Standard BCE: 0.71 Dice → Focal Loss + Weighted Dice: 0.91 Dice.

U-Net Medical Focal Loss

Building Production-Grade RAG Systems

2025 Advanced

Deploying retrieval-augmented generation for accurate, grounded QA in local environments without hallucination.

Challenge: Basic RAG hallucinates when retrieval fails. Production systems need reliability metrics.

Solution: Semantic chunking + Hybrid retrieval (BM25 + dense embeddings) + Reranking layer + Confidence scoring.

$Q = 0.4 \cdot MRR(BM25) + 0.6 \cdot MRR(Dense)$

Results: <5% hallucination rate, 320ms latency (retrieval + generation). Deployed on-prem with Llama 2.

RAG LLM Production
Education

Academic Background

2022 – 2026
Bachelor of Computer Engineering
Khwopa College of Engineering, IOE — Libali, Bhaktapur

Completed a comprehensive Computer Engineering curriculum with emphasis on Artificial Intelligence, Machine Learning, and Software Engineering — with hands-on research in intelligent systems and real-world engineering applications.

Completed
2019 – 2021
Higher Secondary Education (+2), Science
Khwopa Secondary School — Dekocha, Bhaktapur

Strengthened foundations in mathematics and sciences.

Completed
Recognition

Achievements

2025

Locus Pattern Verse

Advanced to Phase 2 of a national tech competition (Locus 2025) by developing an AI-powered generative system for traditional Nepali rug pattern design.

2025

CodeYatra Hackathon

Participated in a national-level innovation hackathon, focusing on AI-driven problem-solving solutions.

2022–23

Hult Prize Nepal

Reached Campus Semi-Finals of Hult Prize 2022–23 with a cross-functional campus team, presenting a social-impact startup concept to a panel of industry judges.

Credentials

Certifications

DataCamp
AI Fundamentals
2025
DataCamp
Data Skills for Business
2025
DataCamp
EU AI Act Fundamentals
2025
DataCamp
Understanding Data Topics
2025
Cisco Networking Academy
Python Essentials 1
2024
Cisco Networking Academy
Introduction to Cybersecurity
2024
Cisco Networking Academy
Introduction to IoT
2024
Programiz Pro
Learn HTML
2024
Programiz Pro
SQL Basics
2024
Programiz Pro
Learn Python Basics
2024
Contact

Let's Connect

Open to AI/ML internships, research collaborations, and full-time engineering roles. Let's build something meaningful.

Email
sahilduwal@gmail.com
↗
LinkedIn
Sahil Duwal
↗
GitHub
Sahahahil
↗