Data lived in separate systems.
Users needed to combine regional, environmental, demographic, logistical, and historical data before they could act.
Product Design Internship · Deloitte US · Fall 2024
From climate data
to decisive action.
During my Deloitte US internship, I helped design Sage: an AI-powered geospatial decision platform for protecting food and water systems before, during, and after climate emergencies.
Read the decisions ↓
Transforming geospatial data to ensure a safer future for all.00The internship engagement
My internship brief: help Deloitte explore a complex public-sector opportunity.
As part of a multidisciplinary SCADpro internship team working with Deloitte Government & Public Services, I helped investigate how geospatial analysis and AI could support faster, more coordinated decisions around food and water security.
The work moved from a broad client brief through research, product strategy, prototyping, testing, and a final concept presentation. My focus was connecting user needs, system complexity, and public-sector value into a coherent product experience.
What I owned and influenced
Research synthesis
Connected secondary research, benchmarks, and expert interviews to product opportunities.
Product framing
Helped narrow the audience, use case, workflow, and value story.
Experience design
Designed and refined critical flows across maps, AI assistance, and resource planning.
Client communication
Translated design decisions into a clear operational and business narrative.
A key recommendation I helped shape
Do not build another map dashboard. Build a decision environment that connects evidence, action, and people.
01Frame the system
The challenge
A high-stakes problem with no single owner.
Food and water resilience crosses emergency management, agriculture, public health, transportation, utilities, and policy. Each team sees a different piece of the same event.
How might we use geospatial analysis and insight to help government agencies forecast and plan around extreme climate events, supporting decision-making for food and water security?
Technical data did not equal clarity.
Dense maps and reports slowed non-technical decision makers and made cross-department communication difficult.
Every delay increased exposure.
During emergencies, imperfect or delayed information still had to support resource allocation, evacuation, and public communication.
AI could help, but could not become a black box.
Recommendations needed confidence indicators, traceable sources, and human control.
02Narrow the opportunity
Research synthesis
We moved from “government users” to two decision-making modes.
Benchmark analysis and six expert interviews showed that a useful platform had to serve both immediate resource decisions and long-range resilience planning.
Act now.
Allocate food, water, transport, shelters, and personnel while conditions change.
- Needs clear priority and logistics
- Coordinates across departments
- Reports status to leadership
Prepare next.
Model scenarios, safeguard vulnerable communities, and build support for policy and investment.
- Needs historical and predictive context
- Compares scenarios and regions
- Turns evidence into executable plans
Initial caseCalifornia
We chose California to make the system concrete: wildfire risk, drought, agricultural dependency, water stress, and complex inter-agency coordination converge in one state. The product architecture remained scalable beyond it.
What changed because of research
We stopped treating maps as the destination. The emerging opportunity was a workflow that helps users understand an event, evaluate impact, coordinate resources, document decisions, and learn afterward.
03My design work · Explore and test
From broad platform to critical path
We explored the ecosystem, then tested the decisions inside it.
Early concepts included dashboards, data libraries, contacts, predictive actions, and global maps. Testing helped us prioritize the wildfire response flow as the clearest demonstration of the system.
User testing
Could users read risk, ask for help, and move toward action?
We put the prototype in front of users to observe how they interpreted the map, navigated layers, and used the AI assistant. The sessions exposed the need for clearer action hierarchy and stronger trust cues around generated recommendations.
Make impact zones immediately legibleKeep recommended actions beside the evidenceShow confidence and original sourcesConnect response and reporting in one flow
The final direction
Say hello to Sage.
One environment for situational awareness, AI-assisted action, inter-agency coordination, and institutional learning.
01 · Understand
Layer the event without losing the decision.
The map combines fire spread, population, agriculture, water, facilities, and evacuation information. Tiered risk zones communicate current and predicted impact, while controls let specialists inspect the evidence they need.
Design decision
Keep environmental data on the map and decision data in a persistent side panel, so users can move between “where” and “what now” without changing contexts.
02 · Decide
Turn questions into inspectable actions.
Sage’s assistant proposes evacuation routes and resource strategies using the active event context. Recommendations remain connected to the map, confidence level, and source datasets.
Design decision
Treat AI as an analyst, not an authority. Recommendations are explicit, source-linked, and ready for human review before execution.
03 · Coordinate
Bridge insight and logistics.
Resource managers can inspect shelter capacity, calculate food and water requirements, compare transportation options, and coordinate support with agencies such as FEMA.
Product principle
An insight is only valuable if the system helps the team translate it into ownership, resources, and a next action.
04 · Learn
Make every response useful to the next one.
After an event, Sage compiles impact, damage, safety updates, imagery, and actions into a shareable report. This creates a transparent record for evaluation, funding, and future planning.
Systems decision
Connect live response and post-event reporting so documentation is produced by the work, instead of becoming a separate burden afterward.
04How my work fit the wider service
Beyond the critical path
A connected operating model, not a collection of screens.
The final concept extends from emergency response into data stewardship, cross-agency communication, reporting, and long-term resilience.
CoordinateConnection hub
Find the right agency or expert and move directly into the communication tools teams already use.
AccessData hub
Browse, save, filter, and request datasets across agencies while keeping sources visible.
DocumentReport library
Manage ongoing, predicted, and completed events as a shared record of decisions and outcomes.
PrepareResiliency planning
Use historical and real-time evidence to develop food, water, infrastructure, and community strategies.
05Responsible AI
Building confidence
Useful AI must make uncertainty visible.
In high-stakes government work, speed without accountability creates new risk. Sage pairs generated recommendations with confidence indicators and links to the original datasets, keeping the user in control.
Human review before executionVisible confidence levelsTraceable original sourcesEditable reports and actions
06Value communicated to the client
Impact framework
Measure the quality and speed of decisions, not screen engagement.
Sage is a concept, so these are proposed measures for a pilot rather than claimed launch results.
Time to coordinated action
Measure time from alert to an approved, assigned resource or evacuation action.
Analysis and reporting effort
Measure hours spent finding, reconciling, communicating, and documenting data.
Cross-agency handoff quality
Measure duplicate work, missed dependencies, and time to reach the correct owner.
Resources and communities protected
Measure the effectiveness of interventions across food, water, agriculture, and public safety.
Recommended pilot
Start with one California wildfire-response workflow and a small inter-agency team.
Validate whether Sage reduces the time to understand impact, propose a resource plan, verify sources, coordinate ownership, and produce a post-event report.
The internship experience
I learned how design earns influence in a complex engagement.
I worked within a multidisciplinary team and engaged with Deloitte stakeholders, domain experts, and emergency-management perspectives. My computer science background helped me reason about data and AI constraints; my design role focused on turning that complexity into clear product decisions, testable flows, and a persuasive client narrative.
Translated research into product structureConnected interface decisions to operational valuePrototyped and tested the critical wildfire flowPresented decisions within a client-facing team
07What I took from the internship
Clarity is not simplifying the data. It is making complexity usable.
This internship changed how I think about design in high-stakes, multi-stakeholder environments. I learned to move between research depth, product trade-offs, system constraints, and client communication without losing the human decision at the center.
Sage also strengthened my point of view on responsible AI: an interface cannot simply be intelligent; it must help users inspect reasoning, coordinate responsibility, and remain accountable.
Explore more work →