About me

I’m an AI and data engineer and researcher based in Sydney. I build practical systems across machine learning, computer vision, geospatial data, and LLM applications. I enjoy taking a problem from the data and early design through to a working tool that people can use.

I bring three years of computer vision and research engineering experience from South Korea, followed by applied AI and data engineering work in Australia. My recent work includes large-scale geological data pipelines, machine learning models, agentic workflows, and API-based applications. At Deakin University, I completed a fully funded Master of Information Technology by Research and published research at WACV and ECML PKDD. I received the Women in AI Australia and APAC award in the AI in Mining category.

Outside the lab

A few stories from volunteering, hosting and mentoring in Nepal and Geelong. Open any card to read more.

A mic and a New YearThe Lounge · Geelong

I planned and hosted a Nepali New Year celebration for international students and their families, from sorting out the food to taking the mic as we welcomed 2083 Bikram Sambat.

Hosting Nepali New Year in Geelong
Hosting Nepali New Year in Geelong
Around the tableThe Lounge · Geelong

Some weeks I cooked community meals, decorated the centre for Christmas or joined in on craft days. I also helped new international students find their feet at Deakin orientation.

Cooking community meals
Cooking community meals
Welcoming students at Deakin orientation
Welcoming students at Deakin orientation
Decorating The Lounge for Christmas
Decorating The Lounge for Christmas
Craft day at The Lounge
Craft day at The Lounge
Representing The Lounge at the Victorian International Education Awards
Representing The Lounge at the Victorian International Education Awards
Launch night in GeelongNepali Pathshala Geelong

I helped organise the official launch of Nepali Pathshala Geelong and took the mic as MC. I worked on the run sheet and helped keep the evening moving.

At the podium for the Nepali Pathshala launch
At the podium for the Nepali Pathshala launch
Receiving recognition at the Nepali Pathshala launch
Receiving recognition at the Nepali Pathshala launch
Early steps in STEMSmart Cheli · Girls in Physics

I mentored students through Smart Cheli in Nepal. At the Girls in Physics Breakfast at Deakin, I shared how I found my way into STEM and answered students’ questions about where it could take them.

What i'm doing

  • 3D ICON

    AI and Machine Learning

    I build and evaluate machine learning and computer vision systems, from model development to usable applications.

  • AI and Data Science icon

    Data Engineering

    I create reliable Python and SQL pipelines for large, complex and geospatial datasets.

  • Research and Development icon

    LLM and Agent Workflows

    I build LLM and agent workflows, including retrieval-augmented generation, integrations and business process automation.

  • MLOps icon

    Research and Delivery

    I connect research with practical delivery through evaluation, deployment, documentation and communication.

Resume

Experience

  1. AI & Data Engineer

    Earth AI, Sydney, Australia

    May 2026 - Present
    • Large-Scale Data Pipelines: Built reusable Python, SQL and DuckDB pipelines to clean and validate 70+ million geological assay records for analysis and modelling.
    • Geospatial Predictive Modelling: Developed and evaluated mineral-targeting models across 7+ million national grid cells using XGBoost and LightGBM, using Optuna for hyperparameter optimisation and MLflow to track experiments.
    • Spatial Feature Engineering: Built feature-engineering and spatial-labelling workflows for 3D geochemical modelling across 30+ million drillcore records, using XGBoost.
    • Model Integration: Integrated an anomaly detection model into an internal FastAPI application, adding interpretable maps and saved results for later review.
    • Workflow Automation: Developed and tested AI workflows for task tracking and information retrieval, enabling hourly monitoring and automated updates.
    • Investment Series Demos: Built AI demonstrations for Series C fundraising to explain how the technology worked.
  2. AI Engineer

    HyperGen, Geelong, Australia

    Jun 2025 - May 2026
    • Document Intelligence and Food Compliance: Designed and deployed a client-facing visual document verification application for Australian food compliance checks. Used SAM for segmentation, Mistral OCR to extract text from PDFs, and vision-language models (VLMs) for document classification and quality checks.
    • Production Deployment: Deployed container artifacts to Azure Container Apps with SQL, full-text indexing and scalable workers. Built GitHub Actions CI/CD workflows and implemented Microsoft Entra ID access controls and audit logging.
    • Client Feedback and Testing: Updated document verification rules based on client feedback and reran checks after each change. Used Databricks for testing and experimentation.
    • CRM and AI Agent Platform: Designed and deployed a multi-user web app for timesheet and opportunity tracking, with AI agents for CRM workflows and secure access via Cloudflare Tunnel. Integrated Grafana dashboards for opportunity, product and profitability data and shared them with stakeholders.
    • Employee Workflow Automation: Built Copilot Studio agents and Power Automate flows to automate employees’ routine tasks.
  3. Computer Vision Research Engineer

    Angelswing Inc., Seoul, South Korea

    Aug 2021 - Dec 2024
    • Automated Inspection System: Designed and deployed an AI-powered inspection system for aerial survey data, improving defect detection accuracy by 60%. Operationalised with Docker and CI/CD pipelines, enabling automated retraining and monitoring.
    • Geospatial ML Data Pipeline: Designed and deployed an end-to-end ETL pipeline to automatically detect and process aerial Ground Control Points (GCPs) from raw geospatial survey data, reducing manual data processing by 65%. Applied MLOps best practices (MLflow, DVC, CI/CD) for smooth deployment, validation, and reproducibility.
    • Aerial Image Stitching and Data Processing: Built a real-time aerial image stitching and transformation pipeline, achieving 50% faster processing and 40% reduced memory usage by using computer vision algorithms, GIS data, and mathematical optimisation, producing clean, standardised geo-referenced outputs from heterogeneous raw imagery
    • 3D Mapping & Reconstruction: Developed an MVP data pipeline for 3D mesh generation by integrating computer graphics with Structure-from-Motion techniques, enabling richer geospatial data products from raw aerial datasets
    • 3D Visualisation Tool: Created a prototype UAV imagery visualisation system using Three.js and 3D vision techniques, resulting in a 25% increase in user engagement compared to previous 2D tools.
    • Team Leadership: Led a research initiative coordinating a distributed team of engineers, improving project development efficiency by 20% through agile collaboration (90% of quarterly OKR targets met).

Education

  1. Deakin University

    Master of Information Technology (Research)
    Jan 2025 - Nov 2026 (Completed)
    • Awarded a fully funded research scholarship.
    • Research focus: hyperspectral anomaly detection and machine learning for geoscience.
    • Graduate Teaching Assistant, Mar 2025 - May 2026: taught database and SQL tutorials, graded assignments and supported students.
    • Research Assistant: contributed to machine learning research on CEC prediction.
  2. Tribhuvan University

    Bachelor of Electronics and Communication Engineering
    Nov 2017 - Apr 2022
    • First Class Honours; college topper and three-time semester topper.
    • Committee member of the Robotics and Automation Centre; organised workshops on Arduino, KiCad and AI.

Awards

  1. Women in AI Australia and APAC Award, AI in Mining category

    2026
  2. Women in STEM Mentor Recognition

    For volunteering and promoting technology education among young women
  3. Academic Excellence Awards

    College topper and three-time semester topper

My skills

    Key Tech Skills

    • Machine Learning
    • Deep Learning
    • Computer Vision
    • LLM Applications, RAG and Agentic Workflows
    • Geospatial and Remote Sensing Data

    Programming Languages and Frameworks

    • Python
    • SQL and DuckDB
    • PyTorch, TensorFlow and scikit-learn
    • FastAPI and Flask
    • JavaScript
    • LangChain, LangGraph, FAISS and Chroma
    • Prompt engineering and model evaluation

    Cloud Stack

    • Azure AI and Azure OpenAI
    • AWS: S3, Lambda, API Gateway and ECR
    • Google Cloud: Vertex AI and Cloud Run
    • MongoDB and PostgreSQL

    DevOps/ MLOps Stack

    • Docker
    • DVC
    • MLflow and DVC
    • CI/CD and GitHub Actions
    • Kubernetes
    • Copilot Studio and Power Automate
    • Claude API, n8n and Docker Compose

    Soft Skills

    • Problem scoping and solution design
    • Clear communication with technical and non-technical teams
    • Research, evaluation and documentation
    • Presentations and technical training
    • Team leadership and collaboration

Publications

Peer-reviewed conference papers

  1. Noise-Robust Band Selection via Explainability-Constrained Learning for Hyperspectral Anomaly Detection

    ECML PKDD 2026 · Springer Lecture Notes in Computer Science, vol. 16950, pp. 53-70

    Introduces End-to-End Explainability-Constrained Learning (E2E-ECL), jointly learning band selection and anomaly detection while improving robustness to spectral noise.

  2. DMS2F-HAD: A Dual-branch Mamba-based Spatial-Spectral Fusion Network for Hyperspectral Anomaly Detection

    IEEE/CVF Winter Conference on Applications of Computer Vision (WACV 2026)

    Presents a dual-branch Mamba model that learns spatial and spectral features for hyperspectral anomaly detection.

Journal articles and preprints

  1. A Review on Aerial Image Classification Techniques for Urban Planning

    Journal of Advanced College of Engineering and Management, vol. 11, pp. 119-137, 2025

    Reviews deep learning approaches for aerial image classification and their use in urban planning.

  2. Hyperspectral Anomaly Detection Methods: A Survey and Comparative Study

    arXiv preprint, 2025

    Compares hyperspectral anomaly detection methods across 17 benchmark datasets, discussing accuracy, computational cost and practical trade-offs.

  3. Nepali Voice-Controlled Home Automation System

    International Journal of Advanced Engineering, vol. 6, no. 1, pp. 1-11, 2023

    Combines Nepali speech recognition, a CNN-LSTM model and MQTT to control home devices.

Award

  1. Women in AI Australia and APAC Award - AI in Mining

    2026

    Recognised for work applying AI to mining and geological data.

Portfolio

  • finance

    Loan Approval Web Application

    Web development

  • orizon

    Mail System Clone using Django

    Web development

  • fundo

    Human Activity Recognition Application

    Machine Learning

  • orizon

    Real-Time Image Stitching

    Computer Vision | AI | GIS| Georeferencing

  • orizon

    Automatic GCP Detection

    Computer Vision | AI | Devops

  • deepfake

    Deepfake Audio and Video Detection System

    Applications

  • chatbot

    Chatter-Patter: AI-Powered Chatbot

    Machine Learning

  • nepali-voice-automation

    Nepali Voice Controlled Home Automation

    Applications

  • summary

    Social Network App

    Web development

  • task manager

    Text Mining and Recommendation System

    Machine Learning

  • arrival

    Sudoku Solver using AI

    Machine Learning

Blogs