2026 Career Transition

# 2026 Career Transition

AI Summary

Purpose:

• Keep the current job-search direction, public resume/portfolio entry points,

first application targets, and evidence boundaries for role-specific resume work.

Key points:

• Hyunwook is actively preparing for a job change as of 2026-07-16.
• First high-fit application: Remember & Company Data Engineer.
• User-confirmed status on 2026-08-24: the Remember & Company take-home and

first-round interview passed. The application is awaiting the second round; the CTO and head of AI/Data are reported panelists, while the schedule and format remain Needs confirmation.

• User-confirmed status on 2026-08-28: the GS Retail application (the last one

awaiting a result) is rejected. The user's wording was "GS 칼텍스", but only a GS Retail application exists on record, so the rejection is mapped there (see the 2026-08-28 source note); notification date, stage, and channel are Needs confirmation. No application now waits for a result — the only live process is the Remember second round.

• User-confirmed status on 2026-08-18: CJ ENM Mnet Plus, CJ Olive Young, and

Vroong are rejected. CJ Olive Young was already known as a 2026-08-04 document-screen rejection; the notification dates, channels, and stages for CJ ENM and Vroong are Needs confirmation.

• No employer-specific cause was supplied for the new results. The strongest

evidence-backed retrospective is a role-fit ceiling: the three postings center production Kafka/streaming, distributed processing or data-lake operation, and DW/Data Mart evidence, while the candidate's strongest verified evidence is batch pipelines, data integrity, MySQL, Airflow/Kubernetes, and batch CDC.

• For the next resume revision, lead with three direct operating signals, remove

repeated CDC narration, and make stakeholder/decision evidence more visible. Resume quality must remain separate from whether a posting's core requirements are directly evidenced.

• The Remember interview-preparation guide is available as a Korean source

report and PDF under human/reports/ and output/pdf/. A deployable PDF copy is generated under human/reports/assets/ for the LLM Wiki web portal.

• The second-round preparation set now includes a CTO/AI-Data mock interview

and a Korean day-of one-page brief. Unsupported conflict, failure, feedback, title-scope, and business-outcome answers remain explicit Needs confirmation fields rather than fabricated stories.

• The question-to-study map at

human/reports/2026-08-07-remember-interview-study-map.md connects interview question IDs to existing published study chapters without presenting study knowledge as production experience.

• First stretch application: Toss Payments Data Engineer.
• Toss officially allows simultaneous applications across affiliates; Toss Place

Data Analytics Engineer is the recommended additional Toss-family application.

• User-reported outcome on 2026-07-23: the Toss Place Data Analytics Engineer

application was rejected at document screen. A supplied email screenshot shows a 2026-07-22 14:15 notification and says incumbent-role staff reviewed the documents before choosing someone more suited to the role. No competency-specific reason was provided. The retrospective identifies missing direct Snowflake, dbt, formal DW-modeling, analytical-mart, and product-metric evidence as the highest-confidence inference; the previous high-fit estimate is Outdated and over-optimistic.

• The batch-oriented binlog CDC system is production evidence: Hyunwook directly

designed and built it and completed deployment-scenario review. The technical-support team deploys and operates it in approximately ten customer environments; Hyunwook receives operating issues, improves the implementation, and provides technical support.

• An NHN PAYCO Data Engineer package was prepared on 2026-07-23 but has not

been submitted. Direct evidence covers 5-year ETL/pipeline work, Airflow, Python, SQL/model tuning, Linux scripting, standardization, and automation. Hadoop/Spark/Trino/Hive production work, formal OLAP marts/cubes, Kafka/Flume, analytical OLAP engines, and PAYCO domain experience remain explicit gaps.

• Use separate resumes: Remember should emphasize source collection, mapping,

migration, data quality, Airflow, SQL, and documentation; Toss Payments should emphasize Kubernetes/Airflow platform operations, batch reliability, CDC/binlog work, DQ, monitoring, automation, and incident response.

• The canonical public technical-portfolio URL is

https://llm-wiki.hwlabs.dev/portfolio/; resume and portfolio detail links use the same host under /portfolio/items/<slug>.

• Do not claim production Spark, Kafka, or Flink experience. Describe CDC accurately

as batch-oriented customer delivery rather than real-time streaming.

Relevant when:

• Starting resume, career-description, application, gap-analysis, or interview

work for Remember, Toss-family, or NHN PAYCO roles.

Do not read full document unless:

• Selecting a target role, deciding resume emphasis, or checking public-site

hygiene before submission.

Linked documents:

• ../../sources/career/2026-07-16-job-search-context.md
• ../../sources/career/2026-07-16-toss-systems-resume-review.md
• ../../sources/career/2026-07-23-toss-place-dae-application-result.md
• ../../sources/career/2026-07-23-nhn-payco-data-engineer-posting.md
• ../../sources/career/2026-08-18-cj-vroong-application-results.md
• ../../sources/career/2026-08-24-remember-first-interview-result.md
• ../../sources/career/2026-08-24-remember-second-interview-panel.md
• ../../sources/career/2026-08-27-personality-conflict-collaboration-inputs.md
• ../../sources/career/2026-08-28-gs-retail-rejection.md
• ../people/hyunwook.md
• data-platform-systems-engineering.md
• bts.md
• ../../../human/portfolio/index.html
• ../../../human/reports/2026-08-05-remember-data-engineer-interview-prep.md
• ../../../human/reports/2026-08-07-remember-interview-study-map.md
• ../../../human/reports/2026-08-18-cj-vroong-rejection-retrospective.md
• ../../../human/reports/2026-08-24-remember-second-interview-prep.md
• ../../../human/reports/2026-08-24-remember-second-interview-mock.md
• ../../../human/briefs/2026-08-24-remember-second-interview-onepager.md
• ../../../human/reports/assets/2026-08-05-remember-data-engineer-interview-prep.pdf
• ../../../output/pdf/2026-08-05-remember-data-engineer-interview-prep.pdf

Open Questions

• Exact source repository and local path for resume.hwlabs.dev on the other PC.
• Current base salary, target compensation, earliest start date, and notice period.
• Whether a PDF resume will be added; the live resume currently has no PDF URL.
• Exact Toss Place DAE application date. The rejection-email screenshot shows a

2026-07-22 14:15 notification; its timezone is not displayed.

• Any competency-specific employer feedback for the Toss Place DAE rejection.

The general comparative-fit explanation is known, but the concrete cause analysis remains inference rather than official feedback.

• CJ ENM and Vroong rejection notification dates, channels, and selection stages.
• GS Retail rejection: confirm the company mapping (user said "GS 칼텍스"; only

GS Retail is on record), plus the notification date, channel, and stage.

• Any competency-specific employer feedback for CJ ENM, CJ Olive Young, or

Vroong. The current cause analysis is inference from posting-to-evidence gaps.

• Remember & Company second-round interview schedule, duration, format, exact

interviewer names, and whether any additional participant will attend.

Details

Canonical public entry points

• Resume: https://resume.hwlabs.dev/

- User-confirmed: deployed from another PC; source is not in this repository.

• Technical portfolio: https://llm-wiki.hwlabs.dev/portfolio/

- Source of truth for deployment content: human/portfolio/ in this repository.

Live-site check on 2026-08-07:

• https://llm-wiki.hwlabs.dev/, /wiki/, /portfolio/, and representative

/portfolio/items/<slug> pages returned HTTP 200.

• Repository-managed resume data, the canonical ATS template, and portfolio PDF

sources now point to llm-wiki.hwlabs.dev; the deployed resume app still needs its normal publish step before these source changes become public.

• The resume payload has an empty resume_pdf field.

Recommended first applications

#### 1. Remember & Company — Data Engineer (high fit)

Current listing checked on 2026-07-16:

• https://career.rememberapp.co.kr/job/posting/319051

Why it fits:

• Mandatory requirements map directly to verified work: Python, SQL, Airflow,

pipeline design/operations, root-cause analysis of data-integrity problems, and durable documentation.

• Duties align closely with external-source collection, entity identification and

mapping, large migrations/backfills, reconciliation, quality monitoring, and abnormal-condition detection.

• Preferred Kubernetes, data-quality monitoring, MDM/governance, and AI-assisted

automation also map to existing evidence.

• Spark, Kafka, and OLAP-engine experience is preferred rather than the core minimum,

so the current gap does not undermine the central candidacy.

Resume emphasis:

1. More than 80 crawlers on the common Airflow/Kubernetes platform, with explicit

ownership boundaries between shared infrastructure and crawler-specific logic.

1. Cross-ecosystem entity/version/license/vulnerability mapping and reconciliation.
2. Large backfills, missing-data recovery, idempotent processing, and DQ monitoring.
3. DML Broker DB-access boundary and the verified error reduction.
4. Jira/Confluence-based RCA and operational documentation, described publicly as

work tracking and a document hub.

Application-specific requirement:

• Include a concrete Why Remember, not a generic company-culture paragraph.

Current application status (user-confirmed through 2026-08-24):

• The take-home assignment submitted on 2026-07-28 passed.
• The first-round interview was held on 2026-08-12 with the Data Engineer team

lead and one team member, and the user confirmed on 2026-08-24 that it passed.

• The application is now awaiting the second-round interview. The user reported

that the CTO and head of AI/Data will attend. The date, duration, format, exact interviewer names, and any additional participant remain Needs confirmation.

• The first-round preparation artifacts remain available at

human/reports/2026-08-05-remember-data-engineer-interview-prep.md and output/pdf/2026-08-05-remember-data-engineer-interview-prep.pdf.

#### 2. Toss Payments — Data Engineer (stretch, directionally strongest)

Application channel (user-confirmed 2026-07-17): the Remember-hosted posting

• https://career.rememberapp.co.kr/job/posting/215058

(company listed as 비바리퍼블리카, position "[토스페이먼츠] Data Engineer", 경력 무관, 채용 시 마감 — verified live 2026-07-17)

Toss-direct listing of the same role (checked 2026-07-16):

• https://toss.im/career/job-detail?job_id=4679693003

Note: the two postings word the JD differently. The Remember-hosted version mentions 실시간 데이터 처리 and 검색엔진 환경 구성, and lists Java/Scala/Python; the toss.im version is the fuller JD (Spark 필수, Flink, CDC/스트리밍). Review the resume against both before submission.

Why apply:

• The role is the closest match to the desired long-term Data Platform trajectory.
• Verified strengths include Kubernetes/Airflow platform operations, more than 80

batch workloads, DB performance and reliability, DQ, monitoring, automation, incident response, and binlog-based customer data delivery.

• The CDC project is direct design, implementation, deployment-scenario review,

and post-deployment improvement evidence. Deployment and day-to-day operation across approximately ten customer environments belong to the technical-support team, not Hyunwook.

Material gaps:

• No verified production Spark batch/tuning experience.
• No verified production Kafka operations.
• No verified production Flink operations.
• The production CDC system is batch-oriented by design for customer binlog

delivery. It is strong CDC evidence but must not be equated with a mature Kafka/Flink real-time streaming platform.

Resume emphasis:

1. Shared Kubernetes/Airflow platform reliability and capacity/topology work.
2. Batch scheduling, backfill/reprocessing, DQ, and operational automation.
3. Directly designed batch-oriented MySQL/binlog CDC, completed the deployment

scenario review, and improved issues reported from approximately ten customer environments after the technical-support team deployed it.

1. Concrete incident metrics and the design decisions behind each improvement.
2. Clear gap disclosure: production / operations / PoC / technical review / learning.

Decision

• Submit both roles with separate documents.
• Treat Remember as the high-probability, experience-aligned application.
• Treat Toss Payments as the high-upside stretch application for the desired career

direction.

• Do not prioritize Toss Payments Systems Engineer unless intentionally pivoting

from Data Platform toward Private Cloud / systems infrastructure.

Multiple Toss-affiliate applications

Official policy checked on 2026-07-16:

• Toss Community's joining guide explicitly states that candidates may apply to

multiple affiliates at the same time.

• Toss Payments and Toss Place applications can therefore proceed concurrently.
• Keep all employment facts and metrics identical across applications; change only

role framing and project order.

Recommended Toss Place target:

• Data Analytics Engineer: https://toss.im/career/job-detail?job_id=6650841003
• Better current fit than Toss Place Data Engineer because the verified strengths

are SQL, data integrity, standardization, metadata, DQ, mapping, backfill, Airflow, documentation, and cross-team operating rules.

• Gaps that must remain explicit: production dbt/Snowflake, formal DW fact/dimension

modeling, and product-metric design.

• Toss Place Data Engineer is a weaker current fit because it directly requires

Kafka/Kafka Connect and Spark batch/structured-streaming experience.

#### Toss Place DAE application outcome (notified 2026-07-22, reported 2026-07-23)

Verified result:

• Status: rejected (불합격) at document screen.
• Highest confirmed stage: applied (지원완료).
• Notification shown in the supplied screenshot: 2026-07-22 14:15. Timezone:

Unknown because the screenshot does not display one.

• Employer-stated explanation: incumbent-role staff reviewed the documents, and

the decision was framed as choosing someone more suited to the Data Analytics Engineer role.

• Competency-specific employer reason: Unknown. The email does not identify a

technology, experience gap, document issue, or other concrete factor.

Retrospective conclusion:

• The earlier 75-85% fit estimate in the source research is Outdated. It

over-weighted transferable SQL, Airflow, data-quality, standardization, and schema experience while under-weighting the absence of direct Snowflake, dbt, formal dimensional-modeling, analytical-mart, and product-metric ownership.

• The submitted resume was technically readable and visually intact: three A4

pages, successful text extraction, and no clipping or broken glyphs. ATS or PDF failure is not the leading explanation.

• The employer's wording supports comparative role fit as the primary retrospective

axis and makes a purely automatic ATS-only rejection less likely. Because this may be standard template wording, it does not independently prove the exact review process or the specific criterion that determined the decision.

• The resume explicitly stated that dbt and formal DW modeling were being learned.

This was honest but confirmed that central role criteria were not yet backed by production evidence.

• The nine-item portfolio was strongest in MySQL schema/query work, collection

accuracy, ETL, Airflow/Kubernetes operations, monitoring, and infrastructure. It demonstrated an adjacent operational data-engineering profile, not direct Analytics Engineering or DW delivery.

• Secondary risk: the first page carried nine dense achievements, and the third

page was comparatively sparse. The package spent recruiter attention on broad engineering depth without closing the direct-role evidence gap.

Durable rule:

• Classify future Analytics Engineer roles with core Snowflake, dbt, dimensional

modeling, marts, semantic layers, or product-metric requirements as stretch roles until direct evidence exists. Transferable experience must be labeled as transferable rather than counted as full requirement support.

• Keep document quality separate from screening fit. Strong copy and verified

engineering depth do not imply document-screen success when the strongest depth is in an adjacent role.

Current resume transformation

• Existing source reviewed:

/Users/khw/Downloads/김현욱_토스페이먼츠_Systems_Engineer_ATS_이력서 (2).pdf.

• The existing two-page layout is clean and reusable, but its headline and second

page are strongly Systems Engineer-oriented.

• First Data Engineer content draft:

human/resume/toss-payments-data-engineer/resume-draft.md.

• The current detailed cases are the hybrid Kubernetes/Airflow batch-platform

reliability work and the production-deployed batch CDC design plus operations support. Reconciliation remains a quantified experience bullet rather than a full case.

• Five independent CTO review passes ended with an INTERVIEW verdict. The fourth

pass used the Confluence-backed Debezium/Maxwell/Canal comparison, Maxwell FileSink PoC, customer-network constraint, and deprecated-versus-production boundary. The fifth pass revalidated the corrected role boundary: Hyunwook owns design, implementation, deployment-scenario review, and issue-driven code improvement, while the technical-support team owns customer deployment and day-to-day operation. No submission-blocking content issue remains; production Spark batch/tuning and direct Kafka/Flink operation remain material JD gaps.

NHN PAYCO Data Engineer package (prepared 2026-07-23)

• Status: prepared, not submitted.
• Official posting: NHN PAYCO Data Engineer, with required Hadoop DW/mart,

distributed processing, ETL/Airflow, OLAP modeling, SQL/model tuning, and Linux scripting work.

• Verified direct match: 5 years of ETL/data-pipeline work, Airflow with

KubernetesExecutor, Python, SQL/EXPLAIN and schema/index tuning, Linux/Shell, data standardization, monitoring, and operations automation.

• Explicit gaps: Hadoop/Spark/Trino/Hive production work, formal OLAP

mart/cube design, Kafka/Flume, Trino/Druid/ClickHouse analytical environments, Hadoop operations/performance work, and PAYCO payment-domain experience.

• The submitted evidence set deliberately excludes the public DB-engine page

because information_schema.TABLES.TABLE_ROWS estimates could be read as exact measurements. The resume keeps that claim only with the estimate qualifier and an internal source; the selected portfolio instead uses the AWS-to-IDC migration and 767GB backup-automation cases.

• Document quality and screening fit must stay separate. The package can pass

the deterministic quality gate while the Hadoop/OLAP production gap still makes this a stretch application.

CJ-family and Vroong outcomes (user-confirmed 2026-08-18)

Verified status:

No competency-specific employer reason was supplied. The following is an evidence-backed retrospective, not an employer-stated cause:

• CJ ENM's Codex requirement map directly supported only 44.4% of must-have

evidence. Production Kafka streaming, PB/TB distributed processing/storage, S3/GCP data-lake operation, BigQuery/Snowflake, NoSQL depth, and named Data Mart delivery remained gaps. The richer final document improved presentation but could not create those production signals.

• CJ Olive Young passed the document-quality gate at 95.1 with reviewers at

96/95/95 and 70.0% must-have evidence coverage. The unresolved gaps were still central to the posting: Kafka real-time streaming, real-time heterogeneous synchronization, Spark, OGG/Debezium production operation, and GCP. The result demonstrates that a high document score does not remove a role-fit ceiling.

• All three Vroong persona reviews were BORDERLINE. The final package corrected

the Debezium scope, AWS framing, streaming bridge, and DW-learning disclosure, but production streaming and DW/Data Mart modeling remained absent. Business- stakeholder and logistics-domain evidence was also thinner than the technical operations evidence.

• The shared pattern is observational, not causal proof. The directly aligned

Remember application passed document screening, the take-home, and the first-round interview, while the streaming/DW-centered stretch roles did not. Unknown factors include applicant competition, headcount, compensation, and internal hiring decisions.

Durable rules for the next application round:

1. Separate document quality from role fit. A 95+ gate verifies evidence and

presentation discipline, not screening probability.

1. Treat a role as stretch when two or more central must-haves depend on

production Kafka, Spark/Flink, formal DW/Data Mart, or cloud data-lake evidence that is not currently verified.

1. Rebuild the first-page hierarchy around direct operating evidence: batch

pipeline operations, data-integrity verification, and TB-scale MySQL work.

1. Remove repeated CDC narration across summary, career bullets, and cases; use

the recovered space for decision ownership, non-engineering stakeholders, and observed results.

1. Keep scope boundaries exact. Batch CDC, a Maxwell FileSink PoC, and a

Debezium architecture review must not be presented as production streaming.

Tooling prepared for the next step

• jobstack: 16 Codex skills installed for strategy, company research, resume,

career history, application tracking, and interview preparation.

• gstack: Codex host setup completed; generated gstack skills are available from

a new Codex thread.

• Atlassian Rovo connectivity was verified with one cross-product search returning

both Confluence pages and Jira issues.