The most comprehensive bank Interview Answer Review tooling available online.

There are various sampling techniques that can be used depending on the use cases. Random sampling, Stratified sampling, cluster sampling, systematic sampling, snowball sampling

During final stages of testing upgraded version of core banking applications, we encountered a defect which required resolution from development team. After dev deployed the defect, I had very limited time to thoroughly retest and perform regression.

My task was to perform retesting and make sure there are no major defects after the release.

After resolution, based on my prior knowledge of the apps I grouped those testcases that had a potential impact to run automation regression to speed up the testing. While I manually tested the defects with some edge case scenarios. 

I kept the cross functional team updated about the testing progress and the risk involved if testing wasn't completed on time. 

But, I was able to mitigate the risk of not completing the testing on time and major risk of unexpected issues with the help to prior knowledge and proactive communication.

Thus, I learnt from this incident to carefully strategizing the situation and collaborative approach can help to mitigate risks.

Assuming the events are independent, the probability of them not being the same in 2 picks = P(RB) + P(BR)

P(RB) = P(R)*P(B) = 4/6 * 2/6 = 8/36 = 0.22

P(BR) = P(B)*P(R) = 0.22

P(not the same in 2 picks) = 0.22 + 0.22 = 0.44

S（Situation）：在2024年9月底，我所在团队灰度上线了一个重要新功能。上线不久后，我们注意到产品的 NPS 分数出现了明显下滑。我的老板非常担心这个变化是由我们新功能引起的，因此建议我们暂停功能的进一步上线。

T（Task）：考虑到我们灰度期间已经收到不少正向的用户反馈和行为数据，我担心贸然暂停功能会影响节奏，并可能误判问题根源。我希望找出 NPS 下滑的真正原因，并用更稳妥的方式推进问题解决。

A（Action）：我首先分析了 NPS 下滑的时间节点，发现其发生在我们功能上线之前。结合用户评价存在滞后性的特点，我判断两者未必直接相关。随后我深入分析了近期的低分评价和客服 ticket，发现主要问题集中在打印出错、页面崩溃等历史遗留的功能性 bug。
我将这些发现整理成文档分享给老板，建议我们先优先修复这些问题，并继续观察 NPS 变化。同时，我也请他指出是否存在新功能相关的反馈痛点，以便一并跟进优化。我们快速完成了 hotfix，并验证后上线。

R（Result）：两周内，NPS 分数从 xx 回升到 xx（如能填具体更好），我们功能也顺利推进至全量发布，后续使用量增长了 xx%。老板也认可了我们以数据驱动决策的方式，对我在这件事中的判断力和推进力给予了肯定。

With a major product launch on the horizon, expecting to triple the user load. The infrastructure was non-optimized for demand and incident response was non-harmonious across teams with no set process.

As the SRE Manager, I'd have to implement project changes to infrastructure to accommodate high availability and scalability. I discovered a long-standing issue surrounding incident response and observability that would make it all the more difficult to achieve reliability under pressure and peak traffic.

Therefore, instead of fulfilling the project requirements as presented, I became an advocate for change across the organization. I established an observability platform with 100% stack coverage, created custom SLOs/SLIs/error budgets to enhance performance and reliability visibility, drafted and socialized an incident management playbook across the company and held hands-on trainings for all respective teams. In addition, I advocated for Terraform for Infrastructure as Code to help with deployment time and held FinOps reviews to help eliminate redundant costs in cloud spend.

As a result, we did not only survive the 3X traffic upon product launch with zero critical incidents, but we also achieved 15% saves the following years in cloud costs and 40% faster deployments. Many of the processes I created were adopted as company-wide best practices.