Priyam Shah — Designer

Sift AI

Partnership with Collaborative Earth ↗︎

Opportunity

Sift through sparsely documented information, unpublished data and case studies

Solution

Open-source repository, powered by AI to collate relevant resources and community knowledge

Team

Anusha Thalnerkar, Kirtana Kannan, Rishma Bora, Rochelle Ardersher, Marc O'Brien

Status

Under Development

Opportunity

Forest regeneration practitioners face an overwhelming task of sifting through sparsely documented information, including unpublished data, policy documents, and case studies.

solution

Open-source repository which facilitates planning and collaboration, by leveraging AI to collate relevant resources, community knowledge and mapping tools.

community partner

Collaborative Earth

team

Anusha Thalnerkar, Kirtana Kannan, Rishma Bora, Rochelle Ardersher, Marc O'Brien

status

Under Development

duration

6 months

Theory of Change

✳︎

Problem

Forest restoration practitioners must sift through numerous research papers from various sources for each project.

✳︎

We did a thing

Designed an open-source repository that facilitates planning and collaboration, by leveraging AI to collate relevant resources.

✳︎

Impact

Facilitate inter-lab collaboration for improved restoration outcomes.

Research

8+
Interviews

647
Sticky Notes

Actionable
Insights

Research

Fundraise

Implementation

Monitoring

Reporting

􀄶
Study on Redwood cover growth, occurring in 2024

􀄴
Report on Redwood tree growth, produced in 2019

Why did we choose to double down on UX Research?

Affinity Mapping helped in stack-ranking, prioritizing, and filtering insights from interviews. These led to discovering and themes.

✳︎

Sparse and scattered documentation of papers for practitioners to peruse.

“Sometimes we end up with over 20,000 abstracts, and then have to sort through the relevant ones to get the necessary data.”

✴︎︎

Community knowledge plays a major role in the success of the project- it is sparsely documented at present.

“Utilizing the community's expertise enhances storytelling, offering potential for archaeological discoveries or tourism ventures, supplementing income streams.”

❋

Collaboration among practitioners facilitates exchange of information and better outcomes.

“Practitioners work in isolation sometimes; their roles and details of work aren't clear.”

Early Concepts

Why did a LLM-like solution not work?

Upon conducting early usability tests with participants, we learned that a stand-alone GPT-like AI tool felt disjointed with the resource repository.

How much agency shall we provide to users?

Community contribution like volunteer uploads may increase the likelihood of data inaccuracies due to the absence of an information screening process.

What would our MVP?

A library of relevant publications, coupled with powerful ways to search and find resources.

What would be the optimum application of AI?

An underlying and site-wide integration of AI, rather than a chat-bot or conversational interface.

Desigin the heart of Sift — Deep Search

Unpacking the UX behind how we built search features for practitioners

✔︎

After multiple rounds of testing

A search field with technical filters, and an expandable bar for generative insights provided the highest adoption amongst users.

✳︎

Through testing, we learned that

The search bar must run deep into the product, assisting in insight lookup and filters

✳︎

Through testing, we learned that

Integrating a CTA for insight generation under search didn't look as important as it was

✳︎

Through testing, we learned that

Splitting Deep Search and Smart Insights was a bold step – and one in the right direction.

Product Features

A home for everything important

The workspace enables collaboration with a Personal Archive for organizing resources and Collections for teams to share and communicate through notes. Save your most valued resources, and come back at will.

An easier and faster way to research

Sift through research papers, case studies, and other publications. With a standardized format, practitioners now view the most relevant content first.

UI Cards

Related and similar content shows up as cards. Tags are indicative of the type of publication.

A distinct look

Smart Insights UI elements were intentionally designed to look different with custom gradient for discoverability and ease of glance.

An expert AI model

With over 100,000 data sets already indexed, practitioners can harness Machine Learning to find relevant data in seconds.

Practitioners' Workspaces

Just like real folders, apply tags to categorize saved files – and find them quickly. The workspace enables collaboration with a Personal Archive for organizing resources and Collections for teams to share and communicate through notes. Save your most valued resources, and come back at will.

A distinct look

Smart Insights UI elements were intentionally designed to look different with custom gradient for discoverability and ease of glance.

An expert AI model – trained for the task at hand

With over 100,000 data sets already indexed, practitioners can harness Machine Learning to find relevant data in seconds.

All-new Research

We designed an easier and faster way to go through research papers, case studies, and other publications. With a standardized format, practitioners now view the most relevant content first.

Design System Architecture & UI elements

Modal—Login/Create Account

Modal—Search Filter

Modal—Workspace

Text input

Card—Publication

Card—Sources

Tooltip

Chips—Filters

Toast

Radio Buttons

I ensured the we continually contributed to and made use of a custom design system. Sift evolved fast & often, and it was crucial to make sure changes were consistent, on-brand and scalable for extended use cases and features for the future.

Tool-wide integration of AI meant identifying optimum checkpoints where AI would improve practitioner workflows. I focused on designing two crucial pillars of the Sift.

Deep Search

Find relevant resources faster

Smart Insights

Summarize or pull
insights from resources

Feedback

Finding resources specific to project location was paramount for practitioners, and simply typing in the search location would hinder specificity of location.

↓

Smart Search in action

Generative insights

Insight cards make it easer for practitioners to know if they're looking at the right resource for projects. Here are some explorations.

Throughout testing, Insights cards received thumbs up. However, visual distinction remained a head scratcher. To combat this, I proposed custom fill & accent colors for AI features.

Smart Insights UI

Impact Measurement

The tool is currently in development with the dataset still in training.
Here are some of the anticipated impacts.

✳︎

User Engagement

Total user base, Rate of new users logging in Geographic Distribution

✴︎︎

Number of partnered labs

Quantifying the network, and
building new collaboration ways.

❋

Content Utilization

Number of resources saved,
Number of projects created

User Engagement

Total user base
Rate of new users logging in
Geographic Distribution

Number of
partnered labs

Quantifying the network, and
building new collaboration ways.

Content Utilization
‍

Number of resources saved
Number of projects

Learnings

When in doubt, test it out!

A tool like this has never existed, and so we had no historic data or information to play with. Hence, we turned to user tested to validate and bust our ideas.

Iterate often

If the first version and the final version of Sift met, they wouldn't recogize each other. Iterations, in the right direction and with the right feedback, did make our product simpler, effective and more capable than our original vision.

Teamwork = Dreamwork

Working with this super talented team has been one of the greatest learning experiences during my time in Design. Our meetings were full of "here's an idea", honest critique, and gallons of coffee!