hi, I'm Shreya
A human-centered
visual designer.

the problem

what we learned

• 1

  Users grew up in families where SRH was “impolite” - information came from outside the home.

• 2

  Credibility was hard to verify in English; medical terms lost their cultural context in translation.

• 3

  Users wanted anonymous, encrypted, non-judgmental spaces to ask sensitive questions.

• 4

  Real-life scenarios beat long clinical articles for navigating US healthcare and insurance.

approach

1. 01

   Research

   6 semi-structured interviews with Asian women (18–24) at Cornell, a competitive audit of 8+ SRH apps (Flo, Clue, Nurx, Planned Parenthood) and affinity diagramming on FigJam.

2. 02

   Solution space

   Mapped 55 design ideas across digital, physical and hybrid modes, then narrowed to three tasks: AI Chatbot, Multilingual Health Library, and Play & Learn.

3. 03

   Prototype

   Paper sketches to low-fi wireframes to a medium-high fidelity Figma prototype - each task built as a vertical prototype with complete end-to-end flows.

4. 04

   Evaluate

   Moderated think-aloud usability tests plus a 24-point heuristic evaluation drove three major iterations: a completion page with score, an AI model picker and privacy popup, and translation expanded across every feature.

outcomes

the problem

what we learned

• 1

  Multi-match selection is the highest-friction moment in the flow: long voice lists cause confusion and “get me an agent” behaviour.

• 2

  First-turn intent accuracy is the highest-risk metric: the same misclassification left one participant calm at 4/5 trust and collapsed another to 1/5.

• 3

  Answers without a next step feel incomplete - “tell me the rule” scored far lower than “help me decide what to do”.

• 4

  Short recoveries beat explanations: “I got that wrong, let me try again” outperformed every long apology we tested.

• 5

  Across 6 competitors (Asurion, AppleCare, Lemonade, Progressive, Geico, State Farm), 82% of carriers use AI - but mostly back-office, not customer-facing. Rental car coverage is the right first branch.

approach

1. 01

   Discovery

   Competitive analysis across six insurers and device-protection peers, AI disclosure and consent research, and an analysis of eight pages of the Auto IVR call-flow architecture - annotating each node with friction, retry behaviour and containment risk.

2. 02

   Design the gates

   A three-gate eligibility model: branch (supported DNIS route), scenario (rental eligibility, reimbursement, arrangement) and context (caller identified, contract or claim located). If any gate fails, the caller stays in the standard IVR - plus claim-status filtering so denied and closed claims skip GenAI entirely.

3. 03

   Prototype the flow

   A nine-stage “to-be” flow - IVR entry, caller ID, record resolve, silent route, three-gate check, the Bridge, Gen Agent, IVR return, live agent - built as an interactive call-flow prototype with testing scripts and a Wizard-of-Oz tool.

4. 04

   Validate & write the playbook

   15 sessions across two rounds (10 end-to-end, 5 on error handling), then a vendor-ready prompt table: what to say, what to avoid, and why, backed by the A/B data behind each line.

outcomes

the problem

what we learned

• 1

  Energy - not just access - is the hidden barrier to food security.

• 2

  Trust and familiarity outrank speed and convenience for every participant.

• 3

  Predictable schedules and plain language reduce anxiety more than more features do.

• 4

  Human contact turns a transaction into a reason to show up.

approach

1. 01

   Discover

   Market research plus interviews with older adults (60–78), caregivers and community service providers, using contextual inquiry, surveys and affinity mapping.

2. 02

   Define goals

   Food access without apps or accounts; less physical and cognitive strain; dietary flexibility; trust through consistency; social connection; dignity preserved.

3. 03

   Design the service

   Ordering by phone, paper form or in person. Same routes at the same times weekly, backed by printed schedules and reminder calls. A waist-high serving ledge so nobody bends, climbs or enters the truck. Optional seating turns pickup into a community moment.

4. 04

   Evaluate

   Three rounds: provider and older-adult interviews, low-fidelity concept testing, and medium-fidelity testing of the website and truck visuals.

outcomes

the problem

what we learned

• 1

  Pain point - people worry about affording the financial responsibility of a pet.

• 2

  Pain point - homes aren’t pet friendly and deposits are often required.

• 3

  Motivator - fostering is a low-commitment trial run for people considering adoption.

• 4

  Behaviour - most people feel empathetic towards animals but rarely convert that into action.

approach

1. 01

   Frame

   How might we help social impact organizations express their value, and inspire people to take action in order to drive positive change?

2. 02

   Research

   Secondary research on BC shelter capacity plus user interviews, synthesised into pain points, motivators and behaviours - and into Ella, a 28-year-old QA engineer who wants a dog but isn’t ready to own one.

3. 03

   Design

   Wireframes for home, foster animal, application, booking and confirmation pages, layered with a hexagon-led visual system: Fira Sans headings, Lato body, deep navy, sky blue and ember accents.

4. 04

   Test & iterate

   Usability testing produced three fixes: a sticky navigation on scroll, a clearly labelled visitation location on the booking summary, and a clickable progress bar so applicants can move back through steps.

outcomes