.
Lumeir — Identifying Top Talent Using Predictive AI
Lumeir — Identifying Top Talent Using Predictive AI
Designed to cut time-to-fill by replacing keyword screening with predictive, performance-based AI.
Role
Lead Product Designer
Collaboration
Product Designer (2)
Technical Coordinator (1)
Founder (2)
Duration
2 months
Role
Lead Product Designer
Collaboration
Product Designer (2)
Technical Coordinator (1)
Founder (2)
Duration
2 months
Role
Lead Product Designer
Collaboration
Product Designer (2)
Technical Coordinator (1)
Founder (2)
Duration
2 months
Lumeir is a talent-matching AI startup focused on improving the efficiency of hiring while preserving human judgement.
Lumeir is a talent-matching AI startup focused on improving the efficiency of hiring while preserving human judgement.
Lumeir is a talent-matching AI startup focused on improving the efficiency of hiring while preserving human judgement.
This image is captured from Lumeir's homepage.
They challenged our team to envision blue-sky, future-forward concepts that push boundaries, and to create something that doesn't currently exist in the market.
They challenged our team to envision blue-sky, future-forward concepts that push boundaries, and to create something that doesn't currently exist in the market.
Building off of Lumeir’s existing business model that emphasizes candidate matching, I refined our project focus to prioritize the front-end of the recruiting workflow.
Competitive Analysis
Existing platforms prioritize workflow acceleration, but reduce candidates into over-synthesized summaries, discarding context & transferable skills.
To ensure that our solution has no overlap with existing industry tools, I researched main competitors (Humanly, Juicebox, Ashby, Greenhouse) to understand the current market landscape.
Competitive Analysis
Existing platforms prioritize workflow acceleration, but reduce candidates into over-synthesized summaries, discarding context & transferable skills.
To ensure that our solution has no overlap with existing industry tools, I researched main competitors (Humanly, Juicebox, Ashby, Greenhouse) to understand the current market landscape.
Interviews & Card-Sort Workshop
I interviewed nine recruiters, then transcribed and thematically coded the data, using affinity mapping to surface key industry pain points. Using those pain-points, I facilitated a card-sorting workshop. This guided recruiters through prioritizing their pain points, uncovering what mattered most and why.
Interviews & Card-Sort Workshop
I interviewed nine recruiters, then transcribed and thematically coded the data, using affinity mapping to surface key industry pain points. Using those pain-points, I facilitated a card-sorting workshop. This guided recruiters through prioritizing their pain points, uncovering what mattered most and why.
Research Findings
Optimizing hiring process time is priority.
The main performance metric for recruiters is 'Time-to-fill': the time it takes from a job requisition approval to accepting the job offer. Factors including manually reviewing candidate resumes and recruiter & hiring manager misalignment slow this process down.
Resumes aren't enough when evaluating candidates.
Recruiters voiced that existing ATS systems and AI screening tools overemphasize buzzwords, often dismissing candidates who may have strong transferable skills.
Hesitancy towards AI hiring tools.
Most recruiters expressed wariness of AI’s limitation, particularly around accuracy, the impersonal nature, and the potential dangers of AI replacing human roles.
Research Findings
Optimizing hiring process time is priority.
The main performance metric for recruiters is 'Time-to-fill': the time it takes from a job requisition approval to accepting the job offer. Factors including manually reviewing candidate resumes and recruiter & hiring manager misalignment slow this process down.
Resumes aren't enough when evaluating candidates.
Recruiters voiced that existing ATS systems and AI screening tools overemphasize buzzwords, often dismissing candidates who may have strong transferable skills.
Hesitancy towards AI hiring tools.
Most recruiters expressed wariness of AI’s limitation, particularly around accuracy, the impersonal nature, and the potential dangers of AI replacing human roles.
Future Thinking Design Workshop
To tackle Lumeir’s blue-sky challenge, I directed 2 future-thinking design workshops with recruiters, guiding them through exercises to articulate their pain points, priorities, and aspirational tools. This approach allowed us to co-design with real users to innovative ideas that are grounded in their day-to-day experiences.
Future Thinking Design Workshop
To tackle Lumeir’s blue-sky challenge, I led 2 future-thinking design workshops with recruiters, guiding them through exercises to articulate their pain points, priorities, and aspirational tools. This approach surfaced innovative ideas that are grounded in their real, day-to-day experiences.
FTD Workshop Findings
Candidates that excel post-hire is the main driver for recruiters.
Most recruiters find meaning in their role when the candidates they hire end up becoming high performers, integrating well within the company and within their teams.
Hiring manager expectations are disconnected from the current market.
Their role expectations are heavily based on internal assumptions, overlooking what talent is realistically available. This slows down the time-to-fill due to increased back-and-forth between them and recruiters to hone down on finding their 'perfect candidate'.
FTD Workshop Findings
Candidates that excel post-hire is the main driver for recruiters.
Most recruiters find meaning in their role when the candidates they hire end up becoming high performers, integrating well within the company and within their teams.
Hiring manager expectations are disconnected from the current market.
Their role expectations are heavily based on internal assumptions, overlooking what talent is realistically available. This slows down the time-to-fill due to increased back-and-forth between them and recruiters to hone down on finding their 'perfect candidate'.
Opportunity Area
How might we reduce time-to-fill by optimizing the early-stage hiring process, enhancing candidate evaluation beyond the resume, and design a solution that assists without replacing human judgement?
Opportunity Area
How might we reduce time-to-fill by optimizing the early-stage hiring process, enhancing candidate evaluation beyond the resume, and design a solution that assists without replacing human judgement?
Concept Development
'Reverse Engineering' The Perfect Candidate
The system analyzes post-hire performance to understand what drives success, uncovers patterns in resumes and skills, recommends top candidates, and evolves as it gathers more data.
Solves: Recruiter & hiring manager mis-alignment, manual labor in resume reviewing.
Data-Driven Market Expectations
Utilizing data to compares candidates against the current job-seeking market, helping hiring managers adjust their requirements and make more grounded decisions.
Solves: Disconnect in hiring manager expectations, recruiter & hiring manager mis-alignment
AI Trust = Providing Suggestions > Making Decisions
Optional candidate prioritization features and automated workflows keep humans in control. Resume analysis backed by statistics and predictive data link candidate traits to real-world success.
Solves: Retaining 'human touch' in decision-making, AI skepticism
Tooling Optimization
Seamless integration with major ATS platforms like Workday or Greenhouse are essential given the numerous tools recruiters currently use.
Solves: Time-to-fill hindered by platform-switching
Concept Development
'Reverse Engineering' The Perfect Candidate
The system analyzes post-hire performance to understand what drives success, uncovers patterns in resumes and skills, recommends top candidates, and evolves as it gathers more data.
Solves: Recruiter & hiring manager mis-alignment, manual labor in resume reviewing.
Data-Driven Market Expectations
Utilizing data to compares candidates against the current job-seeking market, helping hiring managers adjust their requirements and make more grounded decisions.
Solves: Disconnect in hiring manager expectations, recruiter & hiring manager mis-alignment
AI Trust = Providing Suggestions > Making Decisions
Optional candidate prioritization features and automated workflows keep humans in control. Resume analysis backed by statistics and predictive data link candidate traits to real-world success.
Solves: Retaining 'human touch' in decision-making, AI skepticism
Tooling Optimization
Seamless integration with major ATS platforms like Workday or Greenhouse are essential given the numerous tools recruiters currently use.
Solves: Time-to-fill hindered by platform-switching
Product Flow
Designing the flow to reduce decisions, without removing them.
I structured the product flow to guide recruiters from ATS & performance data sync, to reviewing AI-ranked candidates and making final decisions. Candidates are scored based on predicted success, with recruiters having full control of the candidate approval. This ensures that recommendations are data-driven and move beyond buzzword matching.
Product Flow
Designing the flow to reduce decisions, without removing them.
I structured the product flow to guide recruiters from ATS & performance data sync, to reviewing AI-ranked candidates and making final decisions. Candidates are scored based on predicted success, with recruiters having full control of the candidate approval. This ensures that recommendations are data-driven and move beyond buzzword matching.
Core System (Front-End & Back-End)
A layered architecture that keeps recruiters in control.
After presenting the concept to Lumeir’s founder, I pivoted to a rapid, prototype-driven approach to meet a tight timeline and validate the concept quickly.
After presenting the concept to Lumeir’s founder, I pivoted to a rapid, prototype-driven approach to meet a tight timeline and validate the concept quickly.
Prototyping
Vibecode prompting to ensure clarity, efficiency, and trust.
With Figma Make, I generated 93 iterations, rapidly turning concepts into testable prototypes to explore & validate key interaction patterns.
Prototyping
Vibecode prompting to ensure clarity, efficiency, and trust.
With Figma Make, I generated 93 iterations, rapidly turning concepts into testable prototypes to explore & validate key interaction patterns.
I grounded each iteration in core heuristics to ensure the system communicates value clearly, supports efficient decision-making, and maintains user trust.
I grounded each iteration in core heuristics to ensure the system communicates value clearly, supports efficient decision-making, and maintains user trust.
I incorporated feedback from Lumeir's founder and user tested with recruiters to refine the design, ensuring it aligned with business goals while addressing real-world hiring workflows.
Refinement
Optimizing for brand, clarity, and scale.
Final Prototype
Final Prototype
Explore the prototype below, or click here for the full screen experience.
I incorporated feedback from Lumeir's founder and user tested with recruiters to refine the design, ensuring it aligned with business goals while addressing real-world hiring workflows. This input allowed me to improve clarity, usability, and overall effectiveness of the system.
Impact
what i'd solve next — lack of data & bias detection
If I had more time, I would design to consider the user flow for organizations that don't have any performance data. This could look like an initial reliance on industry-level patterns across similar organizations, and gradually configuring suggestions as company gathers & inputs evaluation data.
I would also further improve the performance data system to mitigate manager bias and unreasonably harsh evaluations. This could look like evaluation based on an aggregate data system across multiple employees and team members, seeking pattern recognition amongst individuals, encourage performance criteria that prioritize observable signals, and detecting anomalies such as consistent low ratings deriving from a single manager.
Refinement
Optimizing for Brand, Clarity, and Scale
The final iteration considers branding alignment, guidance to assist learning curves, contextual references, and opportunity for scalability.
Refinement
Final Prototype
Play around with the working prototype below! Or click here for the full screen experience.
Impact
Reduced time-to-fill & candidate review time — helping recruiters evaluate more candidates, faster.
The final prototype was tested with users, measuring the potential impact of applying this tool to their workflow. Overall results showed a minimized time spent reviewing candidates, providing value to the recruiters by allowing them to efficiently review high volumes of candidates.
Lumeir founders also expressed strong satisfaction with the outcome, speed of project completion and data visualization methods. After Lumeir refines their job-seeker MVP launch, they plan to implement the 'reverse engineering candidate' to their talent-seeking interface.
Impact
Reduced time-to-fill & candidate review time, helping recruiters evaluate more candidates, faster.
Lumeir's founders were particularly impressed by the speed of delivery and the data visualization approach. After Lumeir refines their job-seeker MVP launch, they plan to implement the 'reverse engineering candidate' to their talent-seeking interface.
Future Considerations
What i'd solve next: lack of data & bias detection.
If I had more time, I would design to consider the user flow for organizations that don't have any performance data. This could look like an initial reliance on industry-level patterns across similar organizations, and gradually configuring suggestions as company gathers & inputs evaluation data.
I would also further improve the performance data system to mitigate manager bias and unreasonably harsh evaluations. This could look like evaluation based on an aggregate data system across multiple employees and team members, seeking pattern recognition amongst individuals, encourage performance criteria that prioritize observable signals, and detecting anomalies such as consistent low ratings deriving from a single manager.
Future Considerations
What I'd Solve Next — Lack of Data & Bias Detection
If I had more time, I would design to consider the user flow for organizations that don't have any performance data. This could look like an initial reliance on industry-level patterns across similar organizations, and gradually configuring suggestions as company gathers & inputs evaluation data.
I would also further improve the performance data system to mitigate manager bias and unreasonably harsh evaluations. This could look like evaluation based on an aggregate data system across multiple employees and team members, seeking pattern recognition amongst individuals, encourage performance criteria that prioritize observable signals, and detecting anomalies such as consistent low ratings deriving from a single manager.