Brix AI Talent Search
Brix AI
Talent Search
An agent-driven recruiting system that help recruiters to find ideal candidates.
An agent-driven recruiting system that help recruiters to find ideal candidates.
An agent-driven recruiting system that help recruiters to find ideal candidates.
OVERVIEW
OVERVIEW
Brix AI is an AI-powered recruiting system that helps recruiters streamline candidate screening through an agent-driven workflow. I collaborated with the team and contributed to the end-to-end design process — from research and concept development to product handoff.
ROLE
ROLE
Product Design Intern(My role in the team)
UX Design Intern
SKILLS
SKILLS
Product thinking / UX design / Motion
Product thinking / UX design / Motion
TIME
TIME
Jun 2025 - Aug 2025
Jun 2025 - Aug 2025
BUSINESS GOAL
BUSINESS GOAL
Build an AI Agentic talent search tool
Build an AI Agentic talent search tool
Vision
VISION
Build an AI recruiting agent that showcases Brix’s vision of using intelligent automation to enhance hiring efficiency.
Value
Value
Demonstrate measurable gains to build investor confidence for the next funding round.
Scalability
SCALABILITY
Develop a standalone, scalable product that recruiters can use instantly — without the need for ATS integration.
THE MAIN PROBLEM
THE MAIN PROBLEM
Recruiters struggle to identify the right candidates efficiently
Recruiters struggle to identify the right candidates efficiently
CURRENT RECRUITING FLOW
😵💫 misaligned Expectations
Recruiters must balance JD requirements, market insights, and internal expectations-making the process complex and error-prone.
😵💫 Endless iterations
Multiple iterations with hiring managers and stakeholders are too time-consuming and labor-intensive, taking 5–15% of total hiring cost.
😵💫 Blind Search
Recruiters search across platforms, rely on keyword filters and intuition, making it hard to spot top candidates.
😵💫 Pain Point 1
Recruiters must balance JD requirements, market insights, and internal expectations-making the process complex and error-prone.
😵💫 Pain Point 2
Multiple iterations with hiring managers and stakeholders are too time-consuming and labor-intensive, taking 5–15% of total hiring cost.
😵💫 Pain Point 3
Recruiters search across platforms, rely on keyword filters and intuition, making it hard to spot top candidates.
IMPACT
IMPACT
Results in three metrics
Results in three metrics
28%
95%
Faster decision-making
Positive feedback from recruiters
28%
Faster decision-making
10Million
10Million
New fundings
New fundings
95%
Positive feedback from recruiters
PRototypes
PRototypes
ICP Analysis: Help users identify the ideal candidate profile (ICP)
ICP Analysis: Help users identify the ideal candidate profile (ICP)
feedback process before first Ranking of candidates
feedback process before first Ranking of candidates
candidates Preview and add to shortlist
candidates Preview and add to shortlist
AI outreach
AI outreach
Design Challenge 1
Design Challenge 1
HMW use AI Agent to help recruiters define and refine Ideal Candidate Profiles (ICP)?
HMW use AI Agent to help recruiters define and refine Ideal Candidate Profiles (ICP)?
After defining the challenge, we explored different flows for recruiters to build ICPs. We chose Proposal A:
Recruiters enter a short JD, and AI generates refined, editable ICPs.
DESIGN EXPLORATION
DESIGN EXPLORATION
With the flow defined, I designed the first version of the experience.
It focused on helping recruiters customize ICPs quickly through a simple, conversational interface.
Initial exploration: Customize ICPs
Initial exploration: Customize ICPs
ITERATIONS
ITERATIONS
We tested this version with our users, and professional recruiters pointed out key gaps in detail and low workflow efficiency.
We found that professional recruiters care deeply about efficiency. Their expectations differ from traditional SaaS products — They want the AI agent to handle most of the work with as few steps as possible..
FINAL DESIGN
Help users identify the ideal candidate profile (ICP)
Help users identify the ideal candidate profile (ICP)
FINAL DESIGN
Design Challenge 2
Design Challenge 2
HMW make the first candidates ranking more accurate and aligned with intent?
HMW make the first candidates ranking more accurate and aligned with intent?
Once the ICP design was approved by both the PM and recruiters, we moved on to candidate search and ranking.
To improve first-round accuracy, we added a quick feedback mechanism, allowing users who generate ICPs with the Mapping Agent to fine-tune results before the full search.
DESIGN EXPLORATION
DESIGN EXPLORATION
Before showing the full ranking, the system presents a small preview of candidates for quick feedback. The agent then learns from this input to dynamically optimize the ranking.
Initial exploration of feedback mechanism
Initial exploration of feedback mechanism
ITERATIONS
ITERATIONS
We iterated after user testing revealed two issues:
Too Limited information for decision-making
Popups disrupt the review flow.
1st iteration
1st iteration
Recruiters still felt the process had too many steps. They were eager to see the ranking results and didn’t want to spend extra time giving feedback.
2nd iteration: Cut down the steps
2nd iteration: Cut down the steps
FINAL DESIGN
Use Feedback to Help users get better results on their first candidates ranking
Use Feedback to Help users get better results on their first candidates ranking
FINAL DESIGN
Design Details- Establish Agent Presence
Design Details- Establish Agent Presence
Make the AI Agent feel central to the workflow
Make the AI Agent feel central to the workflow
INITIAL EXPLORATION
INITIAL EXPLORATION
In our first exploration, recruiters began on the homepage by uploading a job description or setting filters, then entered a flow where the AI agent appeared as a right-side, closable panel.
In our first exploration, recruiters began on the homepage by uploading a job description or setting filters, then entered a flow where the AI agent appeared as a right-side, closable panel.
AI Agents Appeared as a right-side and closable panel
AI Agents Appeared as a right-side and closable panel
However, recruiters noted it felt nothing different from existing recruiting products. Since our goal was to position Brix as an autonomous AI Agent, we needed to shift from a dashboard model to a truly agent-driven experience.
However, recruiters noted it felt nothing different from existing recruiting products. Since our goal was to position Brix as an autonomous AI Agent, we needed to shift from a dashboard model to a truly agent-driven experience.
ITERATIONS
ITERATIONS
We removed all manual filters and placed agents directly at the entry point, making them the core of the interaction instead of remaining them invisible.
The chatbox was moved to the left and redesigned as collapsible but not closable to keep the AI presence consistent.
We removed all manual filters and placed agents directly at the entry point, making them the core of the interaction instead of remaining them invisible.
The chatbox was moved to the left and redesigned as collapsible but not closable to keep the AI presence consistent.
However, a new problem came out in our user testing, we found that users didn’t understand what's the meaning of "Mapping Agent" and "Search Agent", so we renamed them to "ICP Analysis" and "Advanced Search" for better clarity and discoverability.
However, a new problem came out in our user testing, we found that users didn’t understand what's the meaning of "Mapping Agent" and "Search Agent", so we renamed them to "ICP Analysis" and "Advanced Search" for better clarity and discoverability.
Final Design
Final Design
Design System
Build an scalable conversational design system
As Brix AI took on more responsibilities in the recruiting workflow, we needed a scalable design system to unify the interface patterns, ensuring consistency in user experience and efficiency in development.
The system defines how the agent communicates, acts, and delivers results — from chat primitives to structured task outputs.
NEXT STEP
Opportunities I see
Integration with the suite & data migration
Design a standalone ATS and EMS, and reimagine their integration with the AI talent tool—ensuring zero-loss data migration across all three.
Dark mode
Make some explorations on dark mode version to modernize the visual style and give recruiters more flexibility and personalization.
Dark Mode
Dark Mode
AI-driven Product Design
for the next generation of digital products.
Grab a virtual coffee with me ☕
AI-driven Product Design
for the next generation of digital products.
Grab a virtual coffee with me ☕
AI-driven Product Design
for the next generation of digital products.
OVERVIEW
Brix AI is an AI-powered recruiting system that helps recruiters streamline candidate screening through an agent-driven workflow. I collaborated with the team and contributed to the end-to-end design process — from research and concept development to product handoff.
ROLE
UX Design Intern
TIME
June 2025 - August 2025
SKILLS
Product thinking / UX design / Motion
BUSINESS GOAL
Build an AI Agentic talent search tool
VISION
Build an AI recruiting agent that showcases Brix’s vision of using intelligent automation to enhance hiring efficiency.
VALUE
Demonstrate measurable gains to build investor confidence for the next funding round.
SCALABILITY
Develop a standalone, scalable product that recruiters can use instantly — without the need for ATS integration.
PRototypes
ICP Analysis: Help users identify the ideal candidate profile
Feedback Process Before first Ranking of Candidates
Candidates Preview and add to shortlist
AI outreach
Design Challenge 1
HMW reduce the effort of screening candidates so recruiters find the best match faster?
Probelm statement
Overwhelming Profiles
Candidate profiles were too long and cluttered, making it hard to extract key info.
Slow decisions
Evaluating each profile took minutes, delaying the ability to shortlist candidates.
unclear signals
Recruiters struggled to align on what information mattered most, leading to inconsistent judgments.
1st ITERATION
To address these problems, we explored multiple design directions and conducted several rounds of user testing.
2nd ITERATION
We noticed that recruiters felt the highlights on candidate cards were too generic and all in one color. Signals like years of experience or company names didn’t help them make quick judgments.
To address this, we ran additional research and identified the key signals recruiters truly cared about.
FINAL DESIGN
In the first round of user testing, we noticed that recruiters felt the highlights on candidate cards were too generic and all in one color. Signals like years of experience or company names didn’t help them make quick judgments.
To address this, we ran additional research and identified the key signals recruiters truly cared about.
ICP Analysis: Help users identify the ideal candidate profile
DESIGN DETAILS — Establish Agent Presence
Make the AI Agent feel central to the workflow
INITIAL EXPLORATION
In our first exploration, recruiters began on the homepage by uploading a job description or setting filters, then entered a flow where the AI agent appeared as a right-side, closable panel.
However, recruiters noted it felt nothing different from existing recruiting products. Since our goal was to position Brix as an autonomous AI Agent, we needed to shift from a dashboard model to a truly agent-driven experience.
AI Agents Appeared as a right-side and closable panel
ITERATIONS
We removed all manual filters and placed agents directly at the entry point, making them the core of the interaction instead of remaining them invisible.
The chatbox was moved to the left and redesigned as collapsible but not closable to keep the AI presence consistent.
However, a new problem came out in our user testing, we found that users didn't understand what's the meaning of "Mapping Agent" and "Search Agent" so we renamed them to "ICP Analysis" and "Advanced Search" for better clarity and discoverability.
Final Design: Agent-first Entry
We reframed Brix with a persistent chatbox and agent-led workflow, replacing cluttered filters with two clear modes and repositioning the panel to emphasize visual centrality—distinguishing it from SaaS competitors and aligning with our goal of a true AI agentic recruiting tool.
Design Challenge 1
HMW use AI Agent to help recruiters define and refine Ideal Candidate Profiles (ICP)?
DESIGN EXPLORATION
With the flow defined, I designed the first version of the experience.
It focused on helping recruiters customize ICPs quickly through a simple, conversational interface.
ITERATIONS
We tested this version with our users and professional recruiters pointed out key gaps in detail and workflow efficiency.
We found that professional recruiters care deeply about efficiency. Their expectations differ from traditional SaaS products—they want the AI agent to handle most of the work with as few steps as possible..
After defining the challenge, we explored different flows for recruiters to build ICPs. We chose Proposal A—recruiters enter a short JD, and the agent generates refined, editable ICPs.
FINAL DESIGN
Help users identify the ideal candidate profile (ICP)
Initial Exploration: Customize ICPs
Design Challenge 2
HMW make the first candidates ranking more accurate and aligned with intent?
Once the ICP design was approved by both the PM and recruiters, we moved on to candidate search and ranking.
To improve first-round accuracy, we added a quick feedback mechanism
—allowing users who generate ICPs with the Mapping Agent to fine-tune results before the full search.
DESIGN EXPLORATION
Before showing the full ranking, the system presents a small preview of candidates for quick feedback. The agent then learns from this input to dynamically optimize the ranking.
ITERATIONS
We iterated after user testing revealed two issues — limited information for decision-making and popups disrupting the review flow.
Recruiters still felt the process had too many steps. They were eager to see the ranking results and didn’t want to spend extra time giving feedback.
FINAL DESIGN
Use Feedback to Help users get better results on their first candidates ranking
Initial exploration of feedback mechanism
1st iteration
2nd iteration: Cut down the steps
NEXT STEP
Opportunities I see
Refresh the whole suite
The current visual style feels outdated. We need to modernize the design system to match the AI-first direction.
Integration with the suite & data migration
Design a standalone ATS and EMS, and reimagine their integration with the AI talent tool—ensuring zero-loss data migration across all three.
Dark mode & Mobile version
Make some explorations on dark mode version to modernize the visual style and give recruiters more flexibility and personalization.
A mobile version would give them the flexibility to search and connect with candidates on the go — something users mentioned as highly valuable.
DESIGN SYSTEM
Build an scalable AI Agent conversational design system
As Brix AI took on more responsibilities in the recruiting workflow, we needed a scalable design system to unify the interface patterns, ensuring consistency in user experience and efficiency in development.
The system defines how the agent communicates, acts, and delivers results — from chat primitives to structured task outputs.
THE MAIN PROBLEM
Recruiters struggle to identify the right candidates efficiently
CURRENT RECRUITING FLOW
😵💫 Pain Point 1
Recruiters must balance JD requirements, market insights, and internal expectations-making the process complex and error-prone.
😵💫 Pain Point 2
Multiple iterations with hiring managers and stakeholders are too time-consuming and labor-intensive, taking 5–15% of total hiring cost.
😵💫 Pain Point 3
Recruiters search across platforms, rely on keyword filters and intuition, making it hard to spot top candidates.
IMPACT
Results in three metrics
95%
POsitive feedback from recruiters
28%
Faster decision-making
10Million