Background

Punaravartan Foundation is a non-profit organisation dedicated to the repatriation, preservation, and promotion of Cambodia’s and Nepal’s cultural heritage. Its mission is to safeguard and celebrate tangible and intangible treasures, ensuring they are protected, researched, and shared for the benefit of present and future generations. The Foundation supports archaeological research, conservation training, cultural education, and the return of artefacts held abroad, while fostering creativity and scholarship rooted in the history of Nepal and Cambodia.

Currently, the Foundation manages its extensive library of digital assets on Brandfolder, a platform designed for creatives and marketers. With over 15,000 files and no specialised features to support the Foundation’s workflow, the library has become difficult to manage and make sense of. To address this issue, we created a purpose-built platform for the Foundation.

Project UMA

We named our product after Punaravartan Foundation’s project, Project UMA. Uma is a name of Sanskrit origin that means “splendour”, “tranquillity”, and “fame”. It is also associated with the Hindu goddess Parvati. We aim to provide domain-specific features that help the Foundation create associations between different entities and construct relationships between artefacts, locations, and people, while incorporating AI features to speed up its workflow.

Objects

The central entity of our application is the object. It represents a piece of evidence or an artefact of interest. An object can be attached to different pieces of metadata, such as fields and tags, all of which are highly customisable. The metadata associated with an object can also be used to find objects of interest through our powerful search mechanisms.

Project UMA object overview for a Sumerian cuneiform tablet with metadata and collection details

Figure 1.1 Objects

Project UMA objects collection with advanced search filters and artefact cards

Figure 1.2 Collection View

Attachments Upload

Attachments represent files that can be uploaded to and stored on our platform, then associated with objects and events. We support a multitude of file formats, such as images, videos, office documents, 3D files, and emails, that can be viewed directly in our app through intelligent post-processing. We also incorporate AI features such as description generation using large language models and transcription generation to make all content searchable.

Project UMA attachment upload form with a 3D model preview and generated description

Figure 2.1 Attachments Batch Upload

Project UMA attachment viewer previewing a three-dimensional cat model

Figure 2.2 Previewing 3D Files

All entities in our application can be represented as text through descriptions and transcriptions. This is a deliberate design decision that allows users to quickly search through all the data stored in our app using fuzzy search and semantic search.

Project UMA AI semantic search returning gilded bronze artefact results for a natural-language query

Figure 3.1 Semantic Search

Project UMA fuzzy search with match controls and highlighted terms in artefact results

Figure 3.2 Fuzzy Search

Object Tracing

With all associations carefully created, we can visualise all events that have happened to an object over time, including the people involved and the places it has been. We can also use the generated embeddings to find similar objects based on their descriptions and transcriptions.

Project UMA object-tracing timeline and location map for an Aztec obsidian dagger

Figure 4.1 Object Tracing

Project UMA similar-objects tab showing artefacts related to a Southeast Asian bronze crown finial

Figure 4.2 Similar Objects

Project Team

CVWO project team: Yi Hao, Suki, Yilin, Xuan Shan, Henry, and Suvan
  • Hon Yi Hao (Project Lead, Year 2)
  • Apicha Maneerat (Suki) (Developer, Year 1)
  • Zhang Yilin (Developer, Year 1)
  • Ooi Xuan Shan (Developer, Year 1)
  • Henry Lee (Developer, Year 1)
  • Suvan Handa (Developer, Year 1)

Afterthoughts

I am grateful for the opportunity to serve in CVWO again and for the trust placed in me to manage production applications, liaise with clients, and manage a team of brilliant individuals. This is an experience that is hardly found anywhere else in an internship.

– Yi Hao

CVWO gave me the chance to build software while learning that doing so can be meaningful. It has been a challenging and rewarding experience, and I am grateful to have worked alongside such a supportive team.

– Suki

CVWO gave me the opportunity to use the computing skills I learnt in class to serve the community in a structured format, with guidance from computing professionals and professors.

– Yilin

CVWO taught me that simply writing code that works is not enough. I learned that prioritising clean and structured code is essential because it improves both code readability and overall efficiency.

– Xuan Shan

CVWO has provided a rewarding opportunity to learn more about software engineering while helping the community.

– Henry

I joined the CVWO programme expecting to sharpen my technical skills. However, the biggest lesson I learnt was that great engineering is actually about understanding people and building solutions that truly meet their needs.

– Suvan

Acknowledgements

We would like to express our heartfelt gratitude to the following individuals and organisations for their invaluable support and influence throughout our project:

  • GIC for their generous support
  • Prof. Ben Leong (NUS) and Patrick Wang (NUS) for their guidance and mentorship