← All projects
PROJECT 02
Jan 2026 – May 2026
SAMPLESERVE · INTERNSHIP CAPSTONE
Sample
CSI
Digitizing crime-scene evidence collection — so records stay precise when the scene is not.
Interviews
Artifact study
AI-assisted prototyping
Usability testing
Impact
Shipped MVP
0-1 project
ROLE
Product Designer
TYPE
Internship capstone
TIMELINE
Jan – May 2026
TEAM
5-person capstone team
Cover.jpg
Data entry
Guided capture
Required fields and AI prompts at the scene.
Organization
Evidence listing
Primary and secondary relationships, not a pile of forms.
Records
Chain of Custody
Starts at the scene — not after the fact.
01 PROBLEM
problem.md
what was stuck
One missing detail can cost
justice
Officers record evidence in stressful, time-sensitive scenes where attention is scarce. The process still runs on handwritten forms, manual entry, and disconnected systems — which opens the door to missing details, illegible notes, duplicate work, and chain-of-custody gaps.
Scene.jpg
Documented under pressure / paper process.
02 Research
solution.md
Most tools managed evidence
after
it entered the system
We analyzed Axon, Omnigo, and NICE. They were strong once evidence was already in the system — and weak on scene-side capture, where the record actually starts.
Competitive Analysis.jpg
Static forms couldn’t
adapt
to different evidence types
We studied existing evidence forms to see what officers were expected to capture. That became the brief for a dynamic form: repeated fields auto-fill, and evidence-specific fields appear only when they’re relevant.
Artifact Study.jpg
03 SOLUTION
The product had to deliver speed, accuracy, and accountability
at the
same time
Officer needs became three requirements — the direction for the whole experience.
01 Speed
Works fast in the field
Capture evidence quickly under pressure, with as few taps and as little re-entry as possible.
02 Accuracy
Reduces manual errors
Guided capture and dynamic fields catch missing or inconsistent entries before they become problems.
03 Accountability
Every action is traceable
A living chain of custody records who did what, when — prosecutor-ready by default.
AI USE
Once the flow was set, I built the v1 interactive prototype with Claude Code. AI handled scaffolding and implementation so I could focus on flow and edge cases — sketches to a testable product in days, not weeks.
04 DESIGN DETAILS
design.md
Key choices that shaped the
final experience
Dynamic forms over static templates
Different Steps.jpg
Description
Packaging
Media
Lab Exams
Print Label
Tier 1 fields (always required) auto-fill from the case. Tier 2/3 fields show only when relevant — less cognitive load at the scene.
Linking evidences
Linked Evidences.jpg
B123455 · Glass Fragments
PRIMARY
•••
Glass Fragments
09:18 AM · Entryway floor
Draft
DD
1 secondary evidence linked to this item
B123455-A · Blood Sample
SECONDARY
•••
Blood Sample
09:30 AM · Window frame smear
Draft
GN
Testing showed “Parent/Child” confused officers. “Primary/Secondary” matched their mental model on the most-used linking interaction.
AI guardrails
AI recommendation.jpg
✦
AI Recommendations for Blood Sample
Based on evidence type — tap to add
Biology / DNA
Fingerprints
AI guardrails to ensure factually correct evidence collection
chain of custody as a living log
Chain of custody.jpg
Chain of Custody
6
See all
GN
Guy 1 linked Blood Sample to Glass Fragments (Primary)
#B123456 · Apr 15, 2026 06:25 PM
GN
Guy 2 submitted Blood Sample to lab
#B123456 · Apr 15, 2026 06:23 PM
GN
Guy 3 added Blood Sample
#B123456 · Apr 15, 2026 06:23 PM
DD
Dom Dolla added glass fragments
#B123456 · Apr 15, 2026 06:23 PM
DD
Dom Dolla added fingerprint card
#B123456 · Apr 15, 2026 06:23 PM
GN
Guy Nutter created case CAR-2026-0047
#B123456 · Apr 15, 2026 06:22 PM
It updates as evidence is collected, transferred, and handed off — prosecutor-ready by default, not as an afterthought.
05 PRODUCT WALKTHROUGH
walkthrough.md
click through it
Product
walkthrough
We observed what users did, what they said, and where the workflow slowed down — rainbow sheet, quotes, and flow-level friction, not just task completion.
06 OUTCOME
outcome.md
what changed
Knowing where AI helps, and where
human judgement
matters
AI was most helpful when the task was repetitive, technical, or execution-heavy. It helped us move faster and test ideas more tangibly. We were careful not to use it as the decision-maker.
This taught me to use AI mindfully: not to replace thinking, but to reduce busywork and leave more space for better design decisions.
Keep exploring my work!
PROJECTS
Let's Talk
I'm most energized by projects where I can dig into complex problems, collaborate with smart people, and ship things that genuinely improve someone's day.
Reading
Watching
Love Island 🤪
Off Campus
Cooking
ratatoille
Avacado Toast
Biryani
Made with <3 by
Reet
Oberoi
Comment
Reet Oberoi
Open to full-time roles, and interesting conversations about hard design problems.
1