Workbench/resume/projects/proj_39fc41d1dbf22d069896c5d4/session/sess_2a07d5037a7c4fd5b51b0b6fe2462d6b
Event-Driven Fraud Platform deep dive
The transcript keeps interviewer prompts, your answers, coach hints, and the retrieval trace visible beside the structured project context.
- session_id
- sess_2a07d5037a7c4fd5b51b0b6fe2462d6b
- status
- active
- turns
- 3
- latest_trace_hits
- 11
Continue deep dive
Submit the next project answer while keeping the latest interviewer prompt visible above.
Session transcript
Turns are preserved in order so the project narrative can be reviewed, not just improvised once.
interviewer2026-03-24 11:05:38Z
What problem was Event-Driven Fraud Platform solving, and what part did you personally own?
user2026-03-24 11:05:38Z
I owned the event-driven rules engine migration, moved synchronous scoring onto Kafka consumers, and reduced p95 latency by 38 percent while keeping fallback rules for payment checkout.
interviewer2026-03-24 11:05:38Z
Why did you choose Java and Kafka for Event-Driven Fraud Platform?
Coach hints (1)
Hints stay explicit so the session remains a workbench, not a mystery box.
Explain why Java and Kafka fit this project and what it cost you.
Related questions (1)
Related question-bank items are optional supporting context, not the main surface.
Retrieval trace • hybrid • 11 hits
Project deep dive stays anchored on the structured resume_project summary chunk.Slice 5 question-bank retrieval is reused only as supplemental follow-up grounding.resume_project_chunks_synced: 2
source_documentmergedscore 113.0
Question 1: Redis 分布式锁会遇到哪些问题？
Source answer: 要考虑误删、续约、主从切换一致性和超时兜底。
source_documentmergedscore 107.0
Question: Redis 锁怎么回答比较完整？
Answer: 先讲目标，再讲误删、续约、主从切换一致性、超时兜底与监控。
resume_projectmergedscore 100.0
Event-Driven Fraud Platform
Built a real-time fraud scoring pipeline for checkout risk decisions.
Led the async rules engine rollout across payment services.
Reduced p95 decision latency by 38% after stream redesign.
Tech stack: Java, Kafka, Redis, MySQL
source_documentmergedscore 94.0
Question 1: Redis 分布式锁为什么要续约？
Source answer: 因为业务执行时间可能超过锁超时，需要防止锁提前释放。
question_itemmergedscore 62.0
Kafka 重试和死信队列怎么回答？
question_itemmergedscore 62.0
Redis 分布式锁会遇到哪些问题？
question_itemmergedscore 60.0
为什么 ThreadLocal 在线程池里要手动清理？
question_itemmergedscore 59.0
Redis 锁怎么回答比较完整？
source_documentmergedscore 54.0
Question: Kafka 重试和死信队列怎么回答？
Answer: 先讲重试目标，再讲幂等、退避、死信隔离和监控补偿。
question_itemmergedscore 53.0
Redis 分布式锁为什么要续约？
question_itemmergedscore 50.0
ThreadLocal 会导致什么问题？
Project context
Project context stays visible while the conversation evolves.
Built a real-time fraud scoring pipeline for checkout risk decisions.
JavaKafkaRedisMySQL
Led the async rules engine rollout across payment services.
Reduced p95 decision latency by 38% after stream redesign.
Latest prompt
The latest interviewer question stays easy to revisit before answering.
Why did you choose Java and Kafka for Event-Driven Fraud Platform?