JobAid AI
JobAid AI is an AI-powered web application that helps job seekers improve their resumes and cover letters before applying for jobs. Instead of relying on generic templates or manual reviews, the platform analyzes uploaded documents, identifies opportunities for improvement, and generates tailored recommendations to increase a candidate's chances of securing interviews. Beyond document analysis, the product was envisioned as a career companion—one that makes professional writing more accessible, especially for users who may not have access to career coaches or recruitment expertise. As the Lead Product Designer, I led the end-to-end design process, transforming an early-stage product concept into a cohesive digital experience. My work included product strategy, user experience design, interface design, interaction design, design systems, developer collaboration, and the marketing website.
Client
Jobaid AI
DELIVERABLES
BRAND DIRECTION UI/UX DESIGN
Year
2025
Role
Creative Direction
Overview
Job hunting can be surprisingly difficult. You can spend hours searching for a role that feels like a perfect fit, carefully reading through the job description, comparing the requirements with your experience, rewriting your CV, and then spending even more time trying to write a cover letter that sounds professional without sounding generic. And after all of that, there is still that uncomfortable question: Did I actually present myself well enough?
That uncertainty was the starting point for JobAid AI.
JobAid AI is an AI-powered platform designed to help job seekers improve and tailor their resumes and cover letters for specific job opportunities. The initial idea was straightforward: instead of forcing people to repeatedly guess what recruiters, hiring managers, or applicant tracking systems might be looking for, the product could analyse their existing documents against a specific opportunity and help them understand where their application was strong, where it was falling short, and what they could do differently.
But as I started working through the product, I realised that the problem was bigger than simply creating another AI resume checker. The real opportunity was in helping people make better decisions about their applications.
Think about a product designer applying for a Product Designer role. They might have three years of experience, have worked on fintech products, conducted user research, collaborated with developers, and shipped complex mobile experiences. On paper, they may be perfectly qualified. Yet if their CV focuses heavily on visual design and barely communicates product thinking, the application may fail to communicate the depth of their experience.
The problem isn't necessarily a lack of experience. It's a communication gap. And that became the problem I wanted JobAid AI to solve.
When we first looked at the problem, it would have been easy to say, "People need better resumes." But that wasn't really the problem.
The bigger problem was personalisation at scale.
Imagine someone applying for ten jobs in a week. They might have one resume that represents their career reasonably well, but every company is looking for something slightly different. One role might emphasise product strategy, another might care more about research, while another might prioritise leadership or a particular technical skill. The applicant may actually possess all of these experiences, but communicating the right ones for every application requires time and effort.
That's where the process starts breaking down.
People know they should tailor their applications, but tailoring every resume and cover letter manually is difficult to sustain. Our research reflected this behaviour. Users recognised the importance of personalising their applications, but the process could feel overwhelming and time-consuming.
This created two connected problems for us to solve. For the job seeker, how could we make a highly personalised application possible without turning every application into another long editing exercise? And for JobAid AI, how could we use AI to remove the repetitive work without taking away the user's control over their own application?
That second question became particularly important to me. We weren't trying to build a system where someone uploads a resume, presses a button, and blindly accepts whatever the AI produces. The product needed to assist the user rather than replace their judgement.
The principle became simple: automate the repetitive work, but keep the human in control.
My Role & Responsibilities
I worked on JobAid AI as the sole Product Designer and Creative Director, which meant I had the opportunity to influence the product from multiple angles rather than coming into an already-established design system and simply producing screens.
I was responsible for the UX strategy, information architecture, user flows, wireframes, interaction design, high-fidelity UI, design system, brand direction, marketing website and collaboration with developers during implementation.
Working as the sole designer also meant my involvement extended far beyond designing individual screens. I worked with stakeholders to shape the product direction, translated early ideas into user flows and wireframes, established the visual language, designed the web application and marketing website, built the design system, tested prototypes, and worked closely with developers and the AI engineering side throughout the process.
Being the only designer meant there wasn't another person responsible for "UX" while I focused on "UI." I had to think about both at the same time. One moment I could be questioning whether a feature actually belonged in the MVP; the next I could be restructuring a user flow in Figma, reviewing it with stakeholders, and then discussing implementation details with engineering.
That gave me a much broader view of the product. Instead of asking only, "What should this screen look like?", I constantly had to ask, "What problem are we solving here, what does the user need to understand, what does the business need to achieve, and what is technically realistic?"
That became one of the most valuable parts of the project because it pushed me to think about the product as a system rather than a collection of screens.
Research & Discovery
One of the first things I needed to do was bring the different perspectives around the product into alignment. JobAid AI sat at the intersection of several moving parts: the business opportunity, the user's job-search experience, the capabilities of AI, and the technical realities of building the product.
It would have been easy to jump straight into Figma based on the initial product idea, but doing that would have meant designing around assumptions. Instead, I worked with stakeholders to clarify what we were actually trying to build, which problems were most important to solve first, and what the initial product needed to prove.
That conversation helped narrow the early experience around areas where JobAid AI could provide immediate value: resume analysis, job matching, and cover-letter assistance, rather than trying to solve every part of the job-search journey at once.
This alignment became particularly useful later when we had to make decisions about scope. Whenever a new feature or idea came up, we could bring the conversation back to the same question: Does this help us solve the core problem we're trying to validate?
Surveys & Persona
Once we had a clearer understanding of the business direction, I wanted to understand the problem from the user's perspective. We conducted surveys targeted at students and remote workers and held virtual interviews with three participants. These conversations helped shape two primary personas: Aisha, a recent university graduate, and David, a mid-level professional pivoting careers.
Their circumstances were different, but the underlying problem was surprisingly similar.
Aisha needed help understanding how to position herself when she didn't yet have years of professional experience. David had the opposite problem: he already had considerable experience, but because he was moving into a different career direction, his existing experience didn't always map neatly onto the roles he wanted.
That distinction was important.
David might find a product role that asks for experience in areas he has actually worked in, but his resume could still communicate those experiences using terminology from his previous industry. The problem isn't necessarily that David lacks the qualification. The problem is that the connection isn't obvious.
This changed how I thought about what JobAid AI should actually do. The product shouldn't simply correct spelling, improve grammar, or tell someone that their resume "needs work." It should help users understand the relationship between what they already have and what the opportunity is asking for.
In other words, instead of simply saying, "Your resume has weaknesses," the product should help answer a much more useful question: "Here's what this role is looking for, here's what you already have, and here's where the connection could be stronger."
—The patterns we found
Three insights became particularly important as the product took shape.
First, people were already reusing their applications. That behaviour made sense; creating an entirely new application every time simply wasn't sustainable. But it also revealed an opportunity for JobAid AI to reduce the manual work involved in personalisation.
Second, tailoring applications felt like work. Users understood that their applications should be adapted to the role, but the time required to do that repeatedly made it difficult to maintain. The more opportunities someone wanted to pursue, the more painful that process became.
The third insight was perhaps the most important: people wanted AI assistance, but not at the expense of authenticity.
Users were interested in having AI help them, but they didn't necessarily want a completely AI-generated application that sounded nothing like them. They wanted to move faster without feeling like they had handed over ownership of their career story.
That became one of the foundations of the product: JobAid AI should remove effort, not agency.
—Competitive Analysis
Before deciding what JobAid AI should become, I also needed to understand what users were already familiar with. I reviewed products such as Teal, Resume.io, and OpenResume, looking at their features, user journeys, positioning, and approaches to resume creation and optimisation.
The exercise wasn't simply about creating a list of features we could copy. I wanted to understand the expectations users would already bring into JobAid AI. If someone had used a resume builder before, they would already have expectations around templates, editing, recommendations, scoring, and document management. So the question became less about "What features do these products have?" and more about "Where can JobAid AI provide a more meaningful experience?"
One opportunity stood out: context.
A resume doesn't exist in isolation. Its usefulness depends on where the user is trying to take it. A document that works perfectly for one position might not communicate the right experience for another.
That led to an important product direction: the job description should become part of the intelligence behind the experience.
Rather than simply telling someone that their resume could be better, JobAid AI could understand the target role and provide recommendations within that context. That made the AI much more useful because its suggestions were connected to an actual opportunity rather than being generic advice.
Wireframing & Ideation
With the problem becoming clearer, I started translating the thinking into the product experience.
I began with simple sketches and low-fidelity explorations. At this stage, I wasn't interested in making anything beautiful. I wanted to move quickly enough to challenge my own assumptions. I explored different approaches to onboarding, resume analysis, job matching, and cover-letter generation before taking the strongest directions into Figma.
This stage became particularly important because AI products can become complicated very quickly. There are inputs, outputs, recommendations, scores, loading states, generated content, and decisions that the AI can make versus decisions that the user still needs to make.
If all of that complexity is introduced at once, the interface becomes overwhelming.
So instead of simply asking "What screens do we need?", I started asking more fundamental questions: What does the user need to know right now? What can wait? Where should the AI take initiative? And where should the user remain in control?
The wireframes gave us a way to answer those questions before visual design made us emotionally attached to a particular solution. I reviewed the early directions with stakeholders and tested the core concepts before moving into high-fidelity design.
The goal wasn't to produce perfect wireframes. It was to reduce uncertainty early enough that we weren't solving the wrong problem beautifully.
Information Architecture
As the product started taking shape, I organised the experience around four major flows: Resume Upload & Analysis, Job Description Matching, Cover Letter Generation, and AI-Assisted Job Application.
The fourth represented the longer-term direction of the product, while the initial experience focused on validating the core value around resumes, job descriptions, and application materials.
There was an important idea underneath this structure: the resume would become the foundation of the experience.
Rather than repeatedly asking users to provide the same information, JobAid AI could use the document they had already uploaded as a source of context. Once the system understood the user's experience, that understanding could feed into other parts of the product.
The resulting mental model was straightforward:
Resume → Understand the user → Understand the target role → Identify opportunities → Generate relevant recommendations → Create supporting application material.
This made JobAid AI feel less like a collection of disconnected AI features and more like one continuous assistant.
Branding & Visual Design
The visual identity had an interesting problem to solve. JobAid AI is an AI product, but the people using it aren't necessarily interested in AI itself. They're interested in getting a job.
That meant the brand couldn't feel like a futuristic AI experiment. It needed to feel dependable enough to trust with something as personal as a resume while still communicating that there was intelligent technology working behind the scenes.
I directed the visual language around three qualities: intelligent, trustworthy, and approachable.
The interface therefore leaned toward clarity rather than visual novelty. Clean layouts, strong hierarchy, contextual guidance, predictable interactions, and clear feedback states helped make the product feel understandable even when the underlying technology was complex.
Logo icon |
BRAND LOGO |
Color & Typography:
The primary colour was Oxford Blue (#00237A), supported by Skye Blue (#3C9EDB). The darker blue gave the product a sense of stability and professionalism, which felt appropriate for a platform dealing with people's careers and personal information, while the lighter blue introduced a little more energy and optimism.
The combination was intentional: professional enough to trust, but approachable enough to use.
BRAND COLORS IDENTITY |
BRAND COLORS |
—Typography
Typography followed the same philosophy. I used Inter Tight and Plus Jakarta Sans to balance information density with approachability. Inter Tight worked well for areas such as dashboards and analysis screens where users needed to scan information quickly, while Plus Jakarta Sans introduced a softer, more human quality to the broader experience.
The goal was never to make JobAid AI look like a traditional corporate HR platform. It needed to feel modern and intelligent while remaining approachable to someone who might already be frustrated by the job-search process.
PRIMARY COLOR |
SECONDARY COLOR |
Design Execution
The hardest part of designing JobAid AI wasn't actually drawing the interface. It was deciding how much of the intelligence the user should see.
An AI system can perform an enormous amount of work behind the scenes. But exposing all of that complexity doesn't automatically make the product better. Someone uploading their resume doesn't necessarily need to understand how the document was parsed, how the model reached a particular conclusion, or what happened technically between upload and result.
They need to understand the outcome.
So I designed the experience around three simple questions:
What did the system find? Why does it matter? What can I do about it?
That principle influenced the analysis screens, recommendations, editing experiences, and feedback states throughout the product. Instead of trying to prove how sophisticated the technology was, the interface focused on making that sophistication useful.
—Resume analysis
The resume analysis experience was designed to turn what could easily have become an intimidating AI report into something actionable.
A score on its own isn't particularly helpful. If someone receives a 72%, for example, the immediate question is "Why 72%?" and, more importantly, "What am I supposed to do now?"
That meant the interface needed to connect analysis with action. If the system identified a missing skill, weak wording, or an area that could be better aligned with the target role, the user needed enough context to understand the recommendation and decide whether to act on it.
This became even more important during testing, when users told us that some of the initial analysis scores weren't immediately clear. That feedback eventually led to changes in how the analysis and recommendations were presented.
—Designing automation without taking control away
Automation was another area where I had to find the right balance.
JobAid AI could automatically extract information from a user's resume, which reduced the amount of manual work required. But automation can become frustrating when users can't correct it.
Testing showed that while users appreciated the convenience of auto-fill, some still wanted control over what the system was entering.
So rather than treating automation as an all-or-nothing decision, I introduced a manual-fill option alongside it.
It seems like a small interaction decision, but it changed the relationship between the user and the AI. The system could do the heavy lifting, but the user could still step in and say, "That's not quite right."
That was exactly the relationship I wanted the product to establish: AI does the repetitive work; the user makes the final call.
Web Application |
Handoff Snaphot |
Testing & Iteration
The first version wasn't treated as the final answer. I built interactive prototypes in Figma and tested the core experiences with 12 early-access users, focusing on onboarding, resume upload and analysis, cover-letter generation, and document revision history.
The feedback was valuable because it exposed areas where our assumptions didn't completely match how people interacted with the product.
—Onboarding was doing too much
The first onboarding experience introduced too much information before users had experienced the value of the product. Some participants found the process lengthy and overwhelming.
When I looked at it again from the user's perspective, the problem became obvious. Someone who has just discovered JobAid AI doesn't necessarily want to complete a long setup process before understanding why the product is useful.
So I streamlined the onboarding experience and removed unnecessary steps.
The goal wasn't simply to make onboarding shorter. It was to get users to the moment of value faster.
—The analysis score needed context
The initial analysis experience relied heavily on scores, but some users weren't sure what those numbers actually meant or what they should do with them.
I redesigned the feedback structure to provide a clearer breakdown of the analysis and make the recommendations easier to understand.
This reinforced something I continue to apply in my product work:
Information is only useful when people know what to do with it.
—Automation needed an escape hatch
The auto-fill experience reduced effort, but users still wanted the ability to intervene. Introducing manual input alongside automation gave people that control without forcing them to do everything themselves.
It was a small example of a larger principle that emerged throughout the project: good automation should remove friction without removing agency.
Key findings and improvements:
Users loved the auto-fill feature but wanted more control so I added a manual-fill option.
The resume analysis scores were unclear, so I redesigned the feedback layout for better clarity and breakdown.
Some users felt the onboarding was too long and overwhelming, so I streamlined the process by shortening the flow.
Each round of testing brought measurable improvements in task completion time and user satisfaction.
Challenges & Solutions
Like most fast-moving product builds, JobAid AI faced several real-world challenges which are both strategic and executional during the design and handoff process.
1. Simplifying complex AI interactions for non-technical users
Many users were new to AI-powered tools and didn’t fully understand how their input data was being analyzed or scored. Our goal was to offer an intelligent system without overwhelming users with technical jargon or rigid flows.
Solution:
We approached this by designing a clean, conversational interface that guides users step-by-step through uploading their resumes or job links. We replaced complex configuration options with intuitive toggles and smart defaults. Informational tooltips were added at critical points to explain the “why” behind our suggestions without requiring additional user effort. This struck a balance between transparency and simplicity.
2. Optimizing AI prompt engineering for accurate, relevant suggestions
To make JobAid AI genuinely useful, the AI had to not only analyze documents effectively but also generate tailored suggestions for improvement. Initial testing revealed that some outputs were too generic or missed contextual cues from the job descriptions.
Solution:
We invested time in refining our prompt structures, feeding the AI large sets of annotated CVs and job descriptions across different roles and industries. By fine-tuning prompts and layering contextual understanding, we boosted the relevance of AI-generated insights. This work helped us achieve an impressive 95% accuracy rate in analysis and recommendations, ensuring users felt like they were getting expert-level guidance.
Final Outcome
JobAid AI progressed into a private beta, giving early users the opportunity to experience the core product while providing feedback that could continue shaping the experience.
The platform brought together AI-powered resume analysis, job-description matching, resume optimisation, cover-letter generation, and document management, supported by a marketing website and an underlying design system.
The early results were encouraging. The platform recorded 95% accuracy in resume parsing and analysis, while 70% of users completed both the resume and cover-letter flows within a single session. Early feedback also indicated strong perceived value in the quality of the recommendations.
But the more meaningful outcome for me was what the project demonstrated beyond the numbers.
People were willing to incorporate AI into something as traditionally manual and personal as applying for jobs, provided they could understand what the system was doing and retain control over the outcome.
That was the real validation of the product direction.
Key Learnings
AI Needs Human Context to Be Useful
Designing an AI-powered tool isn’t just about generating smart outputs, it's about making those outputs meaningful. We learned that users responded best to suggestions when they were clearly explained and contextually appropriate. This pushed us to focus on making feedback feel personalized, not generic.
Users Want Control, Not Just Automation
Job seekers want AI assistance, but they also want to feel in control of their documents. Features that allowed editing, toggling suggestions, or reviewing feedback in steps saw higher engagement than automated, one-click fixes. Design should empower and not replace the users.Designing for Trust is as Critical as Usability
Trust is everything when you’re asking users to upload personal documents. We found that visual clarity, clean UI, reassuring microcopy, and consistent flows significantly impacted how secure users felt. Building a professional, transparent experience was key to early adoption.Start Simple, Then Expand Based on Real Use
We initially scoped broad features, but narrowed down to solving just two key pain points: improving resumes and cover letters with actionable insights. This focused MVP approach helped us ship faster, gather feedback, and validate demand without overbuilding.Collaborating Across Functions Elevates the Product
Working closely with developers, product leads, and AI engineers allowed us to adapt quickly to technical realities (e.g., prompt tuning, data limitations). Constant iteration and feedback loops made the product stronger and easier to use.Microinteractions Matter More Than You Think
From hover states to loading animations, small UI details made the app feel smoother and more “alive.” Users mentioned these unprompted in feedback sessions — proof that polish can boost perceived quality and trust.Job Seekers Are Emotionally Invested; Design Should Reflect That
Applying for jobs is stressful, and users often feel vulnerable. We learned to be more intentional with tone, hierarchy, and UX writing, using calm, confident language to guide them, not overwhelm them.
Conclusion
JobAid AI started with a relatively simple question:
Can AI make applying for jobs easier?
By the end of the project, I realised the more interesting question was:
How do we make AI genuinely useful to someone who is already under pressure?
That distinction shaped the product. It influenced how we structured the information architecture, how we introduced automation, how we communicated AI-generated feedback, how we approached onboarding, and even how the visual identity was developed.
The project also pushed me beyond traditional UI design. As the sole designer, I had to move between product strategy, research, UX, visual design, AI behaviour, stakeholder alignment, testing, and implementation. I wasn't simply translating requirements into interfaces; I was constantly helping define what the product should be and why.
And that is probably what I'm most proud of about JobAid AI.
Not that I designed a collection of polished screens, but that I helped turn a complex idea—AI-powered career assistance—into something that could feel understandable, useful, and human.
The experience reinforced something I now carry into every product I work on:
The best technology doesn't make people feel like they're interacting with a smarter machine. It makes them feel like they're capable of doing more.