Portfolio

Detailed background, role, results and lessons behind each achievement.

Anonymous Corporate Card Expense Automation B2B SaaS — FlowPay AI-Powered Gyeonggi Local Currency Store Recommendation — Y:Wave AI-Assisted Anesthesia Guidance System — NeuroSight AI-Based Terms of Service Analysis — TCP AI-Powered Personalized Career Mentoring Service — hai AI-Based Mock Interview Service — AInterview AI-Powered Smart Expense Manager — RZi AR-Powered Furniture Recommendation Platform — ARtliving AI Diet Management Service for Diabetics — Meal당 Webcam-Based Posture Correction Service — HuriUP!AI-OCR Mobile Banking for Seniors — SonGeul Support Platform for Socially Isolated Youth — Connect Braille Education Platform for the Visually Impaired — dotori MBTI Community — 16P!ay

Anonymous Corporate Card Expense Automation B2B SaaS — FlowPay

1 / 8

Background

Employees spend 20 min per expense report; accountants process 100+ per month — resulting in ~1,000 hours and ₩80M in annual labor costs per company. As of 2025, only 3 card issuers support easy payment for anonymous corporate cards.
- Interviews with accountants confirmed the biggest pain point: triple repetitive work of 'collect receipts → manually write expense report → enter accounting voucher.'
As a FINNECT Challenge (fintech hackathon) project, performed all roles — PM, frontend developer, and IR presenter — in a 3-person team. (Jun–Aug 2025)

Role

Results & Impact

Won Encouragement Prize at FINNECT Challenge (5th of 102 teams) — rapid hypothesis validation through interviews and pivoting led to this outcome.
Serving as PM, frontend developer, and IR presenter in a 3-person team, I experienced the full B2B SaaS planning cycle: problem definition → solution design → implementation → commercialization strategy.

Lessons Learned

Initially set 'employee convenience' as the Pain Point, but interviews revealed the real Pain Point was 'accountants manually entering vouchers.' Re-defined the solution. Felt firsthand that hypotheses break in the field.
Presenting both technical novelty (FIDO2, Flow ID) and a clear revenue model (enterprise subscription + transaction fees) together earned judges' trust. In B2B, 'business model realism' is a more powerful weapon than 'technology.'

AI-Powered Gyeonggi Local Currency Store Recommendation — Y:Wave

1 / 6

Background

Despite a large base of Gyeonggi local currency users, spending concentrated in large merchants due to difficulty discovering stores — leaving small business owners with minimal benefit.
- As a Likelion 13th cohort project, participated in a national hackathon and planned/developed a service solving the public problem of Gyeonggi local currency store discovery with AI.

Role

Results & Impact

Advanced to the 2nd round of the national hackathon — achieving top 12% among 247 teams.

Lessons Learned

When applying AI in public services, learned that 'UX that makes users trust and act on recommendations' matters more than 'recommendation algorithm accuracy.'
Experienced that translating the macro goal of local economic revitalization into concrete individual user benefits (nearby merchants, discount info) is the core PM competency.

AI-Assisted Anesthesia Guidance System — NeuroSight

1 / 7

Background

78.1% of anesthesia incidents result in death — a high-risk medical procedure. 92.3% of incidents involve non-specialists. A chronic shortage and high turnover of anesthesiologists is the structural root cause.
- Existing B-mode ultrasound relies on subjective interpretation, with no quantitative tissue analysis or real-time anesthetic diffusion tracking. The North American medical AI market is growing rapidly at $58.7B.
GRAFFITI 2025: AI Startup (hosted by KAIST ICISTS) hackathon — served as business idea planner in a 6-person team over 4 days. Designed the concept of applying Barrel Eye's QUS technology to anesthesia.

Role

Results & Impact

Completed GRAFFITI 2025 AI Startup Hackathon (hosted by KAIST ICISTS). Intensively learned medical AI technology, regulation, and market dynamics over 4 days and developed a full commercialization strategy.

Lessons Learned

Medical AI success is determined less by 'technological novelty' and more by 'how accurately you identified field problems' and 'whether the commercialization roadmap is feasible.' Starting from the 78.1% mortality rate data and connecting QUS technology to anesthesia created the planning's persuasiveness.
Translating complex technology concepts into 'who benefits and by how much' was the most important skill in the 4-day hackathon. Converting regulation (FDA/CE) into an OEM revenue stream rather than a risk was the key to demonstrating feasibility.

AI-Based Terms of Service Analysis — TCP

1 / 6

Background

A survey of 131 people revealed that 93% don't read terms of service properly. The SKT SIM hacking incident brought consumer damage from unread ToS to the forefront as a social issue.
- The global legaltech market stands at $3.03B with a CAGR of 7.6–8.7%. Confirmed a blue ocean — no automated real-time ToS analysis service exists in Korea yet.
Existing review services required users to manually copy-paste text, with no tools to understand complex legal language. Concluded that background auto-detection and analysis would be the key differentiator.

Role

Results & Impact

Won campus Ideathon Grand Prize — simultaneously validated the severity of the problem and business viability through 93% survey data and global market size analysis.
Advanced to 2nd round of national hackathon — technical feasibility (BERT/GPT fine-tuning, screen overlay) and specificity of the revenue model received high marks from judges.

Lessons Learned

When solving social problems with technology, analyzing 'why hasn't this existed before' is crucial. TCP gave me experience validating both technical feasibility and market potential while developing an idea into a concrete business model.
Learned that quantitative evidence from 131 survey responses became the powerful message '93%' — the core weapon to persuade judges and investors. A planner's argument only gains persuasive power when backed by data.

AI-Powered Personalized Career Mentoring Service — hai

1 / 6

Background

Identified that job seekers often apply indiscriminately without knowing their strengths or best-fit roles, and face cost and accessibility barriers to 1:1 career mentoring.
- Planned and developed at CHALLKATHON (joint hackathon by HUFS CS Dept × UMC). An environment requiring rapid planning and implementation within a constrained timeframe.

Role

Results & Impact

Completed CHALLKATHON and delivered a working AI career mentoring service.

Lessons Learned

Experienced that focusing on 'minimum features that deliver core value' rather than 'perfect features' within a short hackathon timeframe leads to a higher-quality deliverable. Felt the importance of MVP thinking.

AI-Based Mock Interview Service — AInterview

1 / 5

Background

Research showed 46% of job seekers find interviews the hardest part of job hunting. The project was born from recognizing a critical lack of opportunities to practice in a realistic, repeatable environment.
- Existing prep services only provided question lists. No service combined real-time voice interaction, role-specific questions, and AI feedback.
As the Likelion 12th cohort final project, I led overall planning, design, and frontend development in a 4-person team over approximately 4 months (July–November 2024).

Role

Results & Impact

Successfully completed the voice-based AI interview simulation service and presented it as the Likelion 12th cohort's final project.
Delivered a service capable of generating role-specific interview questions and voice-based interaction through the combination of Azure Speech and GPT fine-tuning.

Lessons Learned

Learned that the key is not just adopting AI technology, but seamlessly integrating it from a UX perspective. Compensating for technical limitations — STT delay, awkward TTS — through UX design was the PM's role.
As a planner, I gained experience balancing 'what's implementable' vs. 'what's ideal' with the dev team. Understanding technical constraints — GPT fine-tuning scope, STT accuracy — and incorporating them into planning was critical.

AI-Powered Smart Expense Manager — RZi

1 / 6

Background

Initiated as a challenge for the K-HTML Hackathon (co-hosted by Seoul Dongdaemun-gu and KHU). Data confirmed that while 21,343 new residents are expected to move into Dongdaemun-gu, traditional market usage among youth stands at just 8.2%.
- Analyzed 6 major pain points of traditional markets: price opacity, complex layout, cash-only payment, no inventory info, unsearchable online, poor parking. Identified 'price opacity' and 'product discovery difficulty' as the core entry barriers.
Established a 3-phase expansion strategy: B2C individual users → B2G local government partnership. Conceived a model to accumulate data through merchant association partnerships and provide local economic revitalization insights to local governments.

Role

Results & Impact

Completed K-HTML Hackathon (Jul 16 – Aug 30, 2025) and delivered a practical community service integrating OCR, AI, and Google Maps.

Lessons Learned

Learned that framing a problem as a user's behavior flow (price compare → discover → purchase) naturally maps features to UX value. Technology (OCR, AI) should be a means to smooth this flow.
Developed multi-stakeholder planning capabilities through designing a service that simultaneously meets the needs of three parties: local government, traditional market merchants, and consumers.

AR-Powered Furniture Recommendation Platform — ARtliving

Background

Confirmed that the inability to preview 'how will this look in my actual space' before buying furniture is a primary cause of high return rates and delayed purchase decisions.
- As a 2025 Spring HUFS H-UP Career Exploration Program project, collaborated with Barrel Eye (AI-based quantitative ultrasound diagnostic software company) to establish the AI technology application direction.

Role

Results & Impact

Won the Grand Prize (Truth Award) at the 2025 HUFS H-UP Career Exploration Program. Judges recognized that the combination of AR and AI recommendation addresses a real consumer problem.

Lessons Learned

Learned that the user problem solved by technology (space visualization before purchase) is more valuable than the novelty of the technology (AR) itself. Technology is a means, not the end goal.

AI Diet Management Service for Diabetics — Meal당

1 / 6

Background

Korea's diabetic population surpassed 6M — 30 years ahead of projections. The Korean Diabetes Association survey confirmed that diet management (44%) is the greatest challenge for diabetic patients.
- Existing diet apps (MyFitnessPal, Noom, etc.) targeted general users. No service integrated medical-grade calorie calculation based on the Korean Diabetes Association's food exchange table with blood glucose management.
As a Likelion 12th cohort project, I led the entire process from problem definition to business model design, UI/UX design, and frontend development.

Role

Results & Impact

Deployed the service at mealdang.vercel.app in a fully usable form. Food-exchange-table calorie calculation and ChatGPT meal auto-generation work correctly.

Lessons Learned

In healthcare, grounding planning in credible sources — academic papers and statistics — simultaneously secures credibility and differentiation.
Through the process of translating the medical concept of a food exchange table into an understandable UX, I experienced firsthand that translating domain knowledge into service is the core PM competency.

Webcam-Based Posture Correction Service — HuriUP!

1 / 5

Background

HIRA statistics confirmed a surge in back disc and spinal disease patients among people in their 20s–40s, worsened by expanded remote work increasing sedentary time.
- Existing solutions like manual therapy (₩300,000–500,000/session) and surgery were expensive and reactive. Identified the absence of a low-cost IT solution for continuous daily posture correction.
As a Likelion 12th cohort ideathon project, served as PM in a 6-person team leading the entire process from problem definition to business model, marketing strategy, and UI/UX prototype. (May 1–16, 2024, 16 days)

Role

Results & Impact

Completed market research, business model, marketing strategy, and UI/UX prototype within 16 days and participated in the Likelion 12th cohort Ideathon.

Lessons Learned

Learned that data-driven problem definition (statistics → Pain Point → solution) is the key to persuasive planning. Without the data on 'surging back disc cases among young people,' conveying the severity of the problem would have been difficult.
As PM, realized the importance of coordinating with dev and design teams on 'technically feasible scope' and adjusting priorities. The core PM role is translating good ideas into realizable specifics.

AI-OCR Mobile Banking for Seniors — SonGeul

1 / 6

Background

Mobile banking usage among seniors 65+ is 53.4% — 41.2%p below non-seniors (95%). Senior population is expected to exceed 10M by 2030 (Statistics Korea 2024), and 71.6% rely on family for financial tasks.
- Bank branch closures and ATM reductions have devastated rural financial access. The leading cause of erroneous transfers is 'account number input error.' Existing apps have high barriers for seniors due to complex authentication and security anxiety.
Submitted to the 2025 KIITI Winter Academic Conference App Development Contest — a solo project covering all phases: planning, AI architecture, and frontend development.

Role

Results & Impact

Won Excellence Award at the 2025 KIITI Winter Academic Conference App Development Contest.

Lessons Learned

Taking the user's existing behavior (handwritten memo) directly as the interface was the key differentiator. For services targeting vulnerable populations, creating a 'structure where users can't make mistakes' matters more than technical completeness.
Just as OCR limitations were compensated with ensemble methods and user verification UI, design thinking that compensates for technical limits through UX led to financial-grade trustworthiness. Turning regulation (FSS mandate) into a B2B revenue stream follows the same logic.

Support Platform for Socially Isolated Youth — Connect

1 / 8

Background

A 2024 Korea Institute for Health and Social Affairs survey found approximately 540,000 socially isolated youth in Korea. Over 80% want to overcome their isolation, yet more than half have never received any support.
- Primary causes of isolation: employment failure (24.1%) and interpersonal difficulties (23.5%). Life satisfaction is about half that of general youth. Social losses are estimated at ₩7 trillion annually.
Existing programs were government-led and face-to-face — paradoxically creating high barriers for isolated youth. Concluded that a non-face-to-face, anonymous, step-by-step approach would be the key differentiator.

Role

Results & Impact

Achieved overall usefulness score of 8.9 (exceeding 8.0 target) in market testing with 26 participants. Weekly challenge received the highest rating (9.4), and AI role-play chat was praised for minimizing aversion to real conversation.
Completed full service planning, UI/UX design, and 40-page frontend development, and led the final presentation and demo.

Lessons Learned

Learned that for services targeting vulnerable populations, accessibility is the core value. The single design principle of 'non-face-to-face and anonymous' decisively lowered entry barriers for target users.
After receiving market test feedback that 38.5% found pricing needed adjustment, I reduced subscription fees — experiencing firsthand that data-driven decision-making improves service quality.

Braille Education Platform for the Visually Impaired — dotori

Background

Framed Braille education access and learning continuity for the visually impaired as a startup camp challenge, designing both social impact and revenue logic.

Role

Results & Impact

Grand Prize at HUFS Startup Camp IR Pitching Competition
Grand Prize at G-RISE Startup Idea Competition

Lessons Learned

Learned that persuasive social-impact pitches pair impact metrics with a credible revenue model.

MBTI Community — 16P!ay

Background

Planned, designed, and developed an MBTI-based community as a Likelion 13th cohort mini project.