Product Design Internship at Microsoft
Reduced 75% of UX issues by proposing to high-level managers and collaborating closely with engineering from ideation to execution for AI-related features and UX improvements that ship quickly.

從 DITL 到微軟，設計師應具備哪些能力？AI R&D UX Design Intern at Microsoft

My Role
Product Designer Intern
Team
Visual Document Intelligent Team
Company
Microsoft, Taipei AI R&D Center
Duration
Mar. 2023 - Sep. 2023
Tools
Figma, Azure AI, Front-end Programming, Microsoft 365
Contributions
While I was mainly working on one product, I also got to try out various tasks and projects, which helped me improve my skills in diverse areas.
New feature development
Created user flows and prototypes to propose and discuss with higher level managers to co-create new features
Cross-functional collaboration
Collaborated with developers for high-quality, fast-shipping  implementation
Exploration chance to improve
Organized user flows visually to support brainstorming and decision making
0-1 product design
Engaged in a new product’s design to facilitate business, engineering and product development
Some numbers...
To learn and grow with the talented team was an amazing journey.
During the internship, I collaborated closely with PM and engineers to ensure product’s implementation and business achievement. Moreover, I pioneered UX consulting and design thinking sharing for the internal team, building the workflow for the the team's first  designer role
8
New Feature Designs
16
UX Improvements
3
Presentations to the Parter and Head from HQ
31
Prototypes
12
Quality Assurances
1K+
Snack hunting 🍪😝
Product Introduction
Document Intelligence Studio, a SaaS product within Azure AI
🖥️
What is Document Intelligence?
An AI service that extracts key data using OCR technology, helping businesses to convert documents into actionable data or train custom models with a no-code approach.
👤
Who are users?
- Companies dealing with lots of documents like invoices, contracts
- IT and Automation Teams using and handling automation solutions
⛳️
What they’re trying to achieve?
- Automate document processing
- Enhance data accuracy
- Support decision-making
Feature 1
Human-In-The-Loop (HITL), a solution to improve custom model accuracy rapidly and intuitively.
- Designed and Iterated: May 2023 - June 2023
- Launched: August 2023
Problem
Labeling, training, and testing processes lack efficiency and integration, causing workflow inefficiencies.
In the original user flow, if the results are not ideal, users have to open the training page again, uploading certain files into training dataset by themselves.
Iterations
The design went through multiple iterations with close collaboration and rapid prototyping.
Proposal 1 - Enable file testing in the labeling page
This proposal integrates the label page and the test page, enabling users to test the model immediately upon receiving it. However, the required development changes and scope would be too large.
Proposal 2 - Enable file testing in the training page
Use “Satisfied/Unsatisfied” as an evaluated word and provide a test summary. However, this concept would impose an additional cognitive burden on users, requiring them to learn a new evaluation framework.
Solution
With the human in the loop flow, the label, train and test loop is tighter and more effective.
- Official announcement ↗
- 800+ manual hours saved per month
- Better quality in document processing
Feature 2
Project Upgrade, a feature to help customers seamlessly transition to the latest API version without any data loss.
- Designed and Iterated: May 2023 - June 2023
- Launched: August 2023
Feature 3
File Gallery, a feature designed to improve the visibility of extensive data sets by creating a list view for files while maintaining consistency throughout the product.
- Designed and Iterated: Apr. 2023 - May 2023
- Launched: August 2023
Mind Learnings
The 7-month internship at Microsoft has been inspiring and motivating.
✨
Embrace technical ambiguity
During the internship, I observed the challenges of designing AI products, given the unpredictable outcomes.
However, the assignment taught me to embrace ambiguity through critical thinking, effective communication, continuous questioning, and quick turnaround.
👐🏻
Connect technology with humans
As the sole designer on the team, I discovered that design serves as the bridge between code, businesses, and humans, fostering actionable and innovative products and services.
Looking ahead, I wish myself keep making impact to leverage the power of AI.
🚩
See the difference
Immersing myself in a leading corporation, I was impressed by my manager's thought-provoking leading, my colleagues' passion for technology, and the culture that empower everyone.
This experience propels me to learn more and become more capable of creating impacts for the world.
Linkedinmedium