Hello, I am

Deepa Khanal

And I'm a

About Me

Crafting Intelligent Solutions

I am an AI Research Engineer passionate about developing intelligent systems that bridge the gap between research and real-world impact.

My expertise spans multimodal AI, document intelligence, and vision-language models. Currently at Vision CoreInfinity, I build production-ready ML systems that solve complex challenges.

0%

Response Relevance

0

Recall@10

0%

Latency Reduction

3+

Years Exp

Multimodal AI
Document Intelligence
ML Pipelines
Deep Learning

B.Sc. Computer Science & IT

Tribhuvan University • 2019 - 2024

DBMSStatisticsMathOSDSAAI/ML

AI Research Engineer

Vision CoreInfinity • Remote, Australia

Experience

Professional
Journey

Building intelligent systems across startups and research labs.

Jan 2025 — PresentCurrent

AI Research Engineer

Vision CoreInfinityRemote, Australia

Leading research and development of multimodal AI systems and vision-language models. Building scalable intelligent solutions for complex enterprise challenges.

Multimodal AIComputer VisionResearch
Jun 2024 — Nov 2024

AI Developer

Applied NLP & Multimodal Systems

SharelookSingapore

Contributed to an LLM-driven educational assistant reaching 88% response relevance. Implemented semantic indexing (Recall@10 = 0.81) and integrated text-to-image generation for educational content.

88%

Relevance

0.81

Recall@10

47%

Latency Red.

LLMsSemantic SearchText-to-ImageMultimodal AI
Sep 2023 — Dec 2023

AI/ML Intern

Data & Generative AI

Prixa TechnologyLalitpur, Nepal

Built preprocessing workflows reducing data cleaning time by 40%. Developed generative AI extraction system for Nepali documents achieving 91% accuracy.

40%

Faster

91%

Accuracy

0.89

F1 Score

PythonPandasGenerative AINLP
Projects

Selected Work

A collection of projects showcasing expertise in machine learning, computer vision, and NLP

Final-Year Research

ASL Recognition System

CNN-based gesture recognition trained on 13,000 images across 26 ASL hand gesture classes with documented failure mode analysis.

92.6%

Accuracy

0.90 F1
PyTorchCNNComputer Vision

Sharelook — Multimodal AI

Multimodal Educational Content

Integrated text-to-image generation for educational content, achieving 85% alignment approval across 1200 evaluated samples.

85%

Alignment

1200 samples
Text-to-ImageGenerative AIMultimodal Systems

ResNet50 Implementation

Transfer Learning Classifier

Fine-tuned ResNet50 for 10-category classification comparing frozen vs fine-tuned strategies with systematic error analysis.

93.4%

Accuracy

+2.1%
TensorFlowResNet50Transfer Learning

Sharelook — LLM Integration

Educational AI Assistant

LLM-driven educational assistant with semantic indexing achieving 88% response relevance and 0.81 Recall@10.

88%

Relevance

MRR 0.67
LLMsRAGSemantic SearchNLP
Skills

Core Expertise

Specialized in building production-ready AI systems with measurable impact

Multimodal AI0%
Document Intelligence0%
Computer Vision0%
NLP & LLMs0%
ML Evaluation0%

Technical Stack

Languages & Frameworks

PythonPyTorchTensorFlowNumPyPandasScikit-learn

Machine Learning

Deep LearningCNNsTransfer LearningNLPEmbeddingsLLMs

Data & Search

Vector DatabasesSemantic SearchOpenSearchRAGData Pipelines

Tools & Cloud

AWSGitDockerStreamlitJupyterMLflow
Contact

Let's work

together

I'm always interested in discussing new research opportunities, collaborations, or innovative AI projects. Feel free to reach out!

Kathmandu, NepalAvailable for remote work