The 2002 movie Catch Me if You Can stars Leonardo Di Caprio as a talented 19-year-old fraudster, who forges millions of dollars in checks while pretending to be a doctor, a lawyer, and even an airline pilot. That Hollywood film now has echoes in the tech industry, in a story which has come to light in Silicon Valley.
2002年的电影抓住了我，如果您可以将莱昂纳多·迪·卡普里奥（Leonardo Di Caprio）饰演，是一名才华横溢的19岁欺诈者，他在假装自己是医生，律师甚至航空公司飞行员时为数百万美元的支票造就了数百万美元。那部好莱坞电影现在在科技行业中回荡了一个故事，该故事在硅谷已经揭示了。
A software engineer in their mid-twenties, based in India, successfully scammed approximately a million dollars annually from tech startups by excelling in interviews, getting hired remotely, and then not doing their assigned work, all while being simultaneously employed by many companies.
一名位于印度的二十多岁的软件工程师每年通过出色的面试，远程雇用，然后不从事分配的工作，从而成功地从科技创业公司骗了大约一百万美元，而许多公司同时雇用了许多公司。
As in ‘Catch Me if You Can’, in this story there’s an unusually talented main character who gets into a dramatic showdown once exposed.
就像在“如果可以的话”中一样，在这个故事中，有一个异常才华横溢的主角，一旦暴露出来就会参加戏剧性的摊牌。
Today’s issue covers what happened, and some learnings from this highly unusual incident:
今天的问题涵盖了发生的事情，以及这一极为不寻常的事件中的一些知识：
Exposé. An engineer named Soham Parekh is accused on social media of working at multiple Y Combinator startups at the same time by one employer, and other workplaces come forward. 18 companies that hired, or nearly hired this engineer. Almost all are recently-founded AI startups. Excuses, excuses… This engineer was hard to contact, and had bizarre reasons as to why. Skyrocketing demand for AI engineers. One thing is clear: demand for AI engineers is through the roof, and capable software engineers who can integrate LLMs are in high demand. How did a dev outside the US get hired by hot Silicon Valley startups? False US work authorization status, Silicon Valley startups desperate to hire, and “business-minded 10x engineers” being in high demand, all helped. Spotting “overemployed” software engineers. Suspicion of ‘overly green’ GitHub profiles, the importance of background checks, and recruitment advice from an ex-Facebook and ex-Uber founder of scaleup hiring business, Dynosaur Labs. Additionally, Digger founder Igor Zalutski confirms that Soham was genuinely among the top 0.1% of software engineers, in terms of both technical skills and business acumen. What does it mean for remote work? Remote work makes being “overemployed” much easier – and AI tooling will make it even more so. Are more in-person interviews and hybrid work patterns a solution?
暴露。一位名叫Soham Parekh的工程师在社交媒体上被指控在社交媒体上由一名雇主同时在多个Y组合创业公司工作，而其他工作场所则挺身而出。18家雇用或几乎雇用这位工程师的公司。几乎所有人都是最近创建的AI初创公司。借口，借口……这位工程师很难接触，并且有奇怪的原因。对人工智能工程师的需求飞涨。一件事很清楚：对AI工程师的需求是通过屋顶，可以集成LLM的有能力的软件工程师需求量很高。美国以外的开发人员如何被热硅谷初创公司雇用？虚假的美国工作授权状况，塞利康谷的初创公司迫切希望雇用，而“有业务意识的10X工程师”需求量很高，这一切都有所帮助。发现“过度就业”软件工程师。怀疑“过于绿色”的github概况，背景调查的重要性以及从前facebook和Pab-uber的招聘建议的招聘建议，以及Scaleup招聘业务的前创始人Dynosaur Labs。此外，Digger创始人Igor Zalutski确认，就技术技能和业务敏锐度而言，Soham真正属于软件工程师的最高0.1％。这对远程工作意味着什么？远程工作使“过度就业”变得更加容易 - AI工具将使之更加容易。更多的面对面访谈和混合工作模式是解决方案吗？
Update: shortly after publishing this article, Soham appeared in the TBPN podcast where he admitted to having worked multiple jobs in parallel since 2022, confirming the allegations in the below article.
更新：发表本文后不久，Soham出现在TBPN播客中，他承认自2022年以来就从事多个工作，证实了以下文章中的指控。
1. Exposé
1。Exposé
Yesterday (2 July), startup founder Suhail Doshi made an accusation: that a software engineer named Soham Parekh was working at several Y Combinator startups at once, and had been doing so for over a year, all while failing to do the work he was hired to do:
昨天（7月2日），创业公司的创始人Suhail Doshi指控：一位名叫Soham Parekh的软件工程师立刻在几家Y Combinator创业公司工作，并且已经这样做了一年多，而这一切都未能完成他的工作：
Initially, the post got a bit of pushback. After all, in California – where most startups which hired Parekh are based – it’s not forbidden to have a second job (aka ‘moonlighting’), as long as it doesn’t overlap with other commitments. Indeed, many leaders founded startups on the side of their main job, and an employer cannot claim intellectual property ownership of a new project which is fully separate from someone’s primary job.
最初，该帖子有点回调。毕竟，在加利福尼亚州 - 大多数雇用Parekh的初创公司所在的初创公司 - 只要它不与其他承诺重叠，就不会被禁止从事第二份工作（又名“月光下”）。的确，许多领导人在主要工作的一边创立了初创公司，雇主不能声称对一个与某人的主要工作完全分开的新项目的知识产权所有权。
What makes this story stand out is the unusually high number of parallel jobs this one dev took on. All together, the combined workload of all these roles was evidently impossible to maintain, and would inevitably lead to questions being asked by individual employers, who wondered why a clearly-talented engineer was unable to deliver their work.
这个故事脱颖而出的是，这个开发人员从事的平行工作异常高。总的来说，所有这些角色的合并工作量显然是不可能维护的，并且不可避免地会引起个人雇主提出的问题，他们想知道为什么一位现年有才华的工程师无法交付他们的工作。
Suhail said his issue was not that Parekh had a side job; it was something more fundamental:
Suhail说，他的问题不是Parekh有一份工作。这是更基本的事情：
Following Suhail’s post, reports from other startups began to emerge, with accusations that they had hired Parekh, and then fired him not long afterward.
在Suhail的帖子之后，其他初创公司的报道开始出现，指控他们聘请了Parekh，然后不久就解雇了他。
He was exposed thanks to the Y Combinator founder community. Doshi went public with accusations after confirming with several other Y Combinator founders who’d had the same experience with Parekh. Garry Tan – president and CEO of Y Combinator – also shared his belief that without the Y Combinator founder community sharing information with one another, Parekh might have not been publicly exposed.
感谢Y Combinator创始人社区，他暴露了他。Doshi与其他几位Y Combinator创始人确认了与Parekh相同的经验后，被指控公开了。Y Combinator的总裁兼首席执行官Garry Tan还表示，他的信念是，如果没有Y Combinator创始人社区彼此共享信息，Parekh可能并未公开暴露。
2. 18 companies that hired, or nearly hired him
2。18家雇用或几乎雇用他的公司
As the drama played out online, things turned bad fast for Parekh. Other companies at which he worked reportedly started taking action, as per Suhail:
随着戏剧在线的发挥，帕雷克（Parekh）的情况变得很糟糕。据报道，他从事的其他公司开始采取行动：
Showdown: founders learned their employee worked at multiple companies at once. Source: Suhail Doshi on X
摊牌：创始人了解他们的员工立即在多家公司工作。资料来源：X上的Suhail Doshi
Many companies went on record about the fact they had employed and quickly dismissed Parekh. Links to companies below lead to their job pages, if they have one. All are hiring for fullstack software engineers or AI engineers.
许多公司都记录了他们雇用并迅速解雇帕雷克的事实。链接到下面的公司会导致他们的工作页面，如果他们有工作页面。所有人都在招聘Fullstack软件工程师或AI工程师。
Playground AI (YC19, AI-first graphics editor.) Hired Soham in 2024 and fired him after a week. Leaping AI (YC W25, building voice AI agents that are human-like and self improve): hired him, then fired him when they found out he also worked elsewhere Fuse AI (YC W25, agentic sales platform for GTM teams): fired him when they found out that Soham was working parallel at Leaping AI Digger (Terraform pull request automation) onboarded him, but terminated him a few days into the job, founder Igor Zalutski told me. The company terminated Parekh when background checks raised red flags. Digger was conscious about the risk of “moving fast” in hiring before background checks completed, but this was a calculated risk. Fleet AI (agent engineering platform). Hired and fired within a week. Lindy (AI assistant platform). Hired and fired a week later. Aced the interview, but was slow with work. Create (creating software with AI): hired him and promptly fired him. Create is an in-person startup. Amusingly, Soham showed up for one day at their SF office, but often had a reason why he could not attend. Animetal (AI platform for infrastructure) Hired in 2022 and fired shortly thereafter. They realized he must be working multiple jobs. Ponder.ai (video AI). Worked at the startup until the news broke, when he was fired. ComfyUI (open source, Node.js-based app for GenAI) hired Parekh for 3 weeks in June. The interview went well, then things started to fall apart.
Playground AI（YC19，AI-First Graphics Editor。）于2024年聘请Soham，并在一周后解雇了他。跳过AI（YC W25，建立人类的语音AI代理商，自我改善）：雇用了他，然后在他们发现他还在其他地方工作时解雇了他（YC W25，GTM团队的代理销售平台）：当他们发现Soham正在发现Soham在Leeh ai degger（Terra digger auptive autape autia tera terra tear auperive autial neip autial terip autip authime terip authim）上的终端时，将他解雇了。伊戈尔·扎卢茨基（Igor Zalutski）告诉我。当背景调查提出危险信号时，该公司终止了Parekh。DIGGER意识到在完成背景检查之前招聘“快速移动”的风险，但这是一个计算的风险。车队AI（代理工程平台）。在一周内雇用并开除。Lindy（AI助理平台）。一周后雇用并解雇。接受采访，但工作缓慢。创建（使用AI创建软件）：雇用他并迅速解雇了他。创建是一个面对面的创业公司。有趣的是，索姆（Soham）在他们的SF办公室出现了一天，但经常有一个原因。Animetal（基础架构的AI平台）于2022年雇用，此后不久就被解雇了。他们意识到他必须从事多个工作。Ponder.ai（视频AI）。在初创公司工作，直到他被解雇时被消息打破。COMFYUI（基于Node.js的开源应用程序）在6月雇用了Parekh 3周。采访进行得很好，然后事情开始崩溃了。
One of the first AI companies to hire Parekh back in 2022. Source: Matt Parkhurst, founder of Animetal on X
最早在2022年雇用Parekh的AI公司之一。资料来源：Matt Parkhurst，Animetal的创始人
Then there are companies which interviewed Soham, but decided against employing him:
然后有一些公司采访了Soham，但决定不雇用他：
A startup advocating for “cheating on everything” also nearly got duped. Cluey is a startup building an AI tool to “cheat on everything”, and as fate would have it, the company’s founder and CEO shared that they were at the second-round interview stage with Parekh. The company shared interview notes taken during their session:
提倡“作弊”的初创公司也几乎被欺骗了。Culuey是一家启动AI工具来“欺骗一切”，正如Fate所拥有的那样，该公司的创始人兼首席执行官分享说，他们正处于与Parekh的第二轮访谈阶段。该公司分享了在会议期间记录的面试：
“Candidate Evaluation: Candidate seen as knowledgeable in React and frontend optimization.
候选人评估：候选人在React和前端优化方面被视为知识渊博。
High-level answers are considered good; some fumbling through details noted.
高级答案被认为是好的；有些人想到了一些细节。
Strong devops and infrastructure experience inferred from GitHub and past projects.
从Github和过去的项目中推断出的强大Devops和基础设施经验。
Candidate’s GitHub is reviewed, showing significant contributions and experience.
审查了候选人的GitHub，显示了巨大的贡献和经验。
Comparison to other candidates in terms of technical strength and fit.
就技术实力和合适而言，与其他候选人进行比较。
Discussion of candidate’s background (e.g., internships at Meta, Waterloo education).
讨论候选人的背景（例如，在滑铁卢教育的Meta实习）。
General consensus that the candidate is a good fit, with a strong,"startery" profile.
一般共识，即候选人非常合适，具有强大的“星tery”特征。
Plan to schedule a follow-up interview for further evaluation.”
计划安排后续采访以进行进一步评估。”
3. Excuses, excuses…
3。借口，借口…
Parekh consistently delayed his supposedly-imminent relocation to the US. Most startups that employed him usually knew he was based in India, but all assumed he wanted to relocate to the US. He assured several employers that he was taking care of his O1 visa (discussed below). For example, here is an email sent to the CEO of Fleet AI:
帕雷克（Parekh）始终推迟了他所谓的迁移到美国的搬迁。雇用他的大多数初创公司通常都知道他是在印度的，但所有人都认为他想搬到美国。他向几个雇主保证，他正在照顾他的O1签证（下面讨论）。例如，这是发送给Fleet AI首席执行官的电子邮件：
“Ideally, I’d love to be in person. However I am also undergoing the renewal process for the current visa. Is it okay if I let you know earliest by this Tuesday if it would be possible to be in person in New York for this [the team working together]. At this point, I’d say it is 80% likely that I should be able to come in person.”
“理想情况下，我很想亲自参加。但是，我也正在接受当前签证的续签过程。如果我在本周二的最早让您知道，如果有可能在纽约亲自与此相处的情况[团队一起工作]。在这一点上，我会说我应该说80％的人可以亲自来。”
Excuses were frequent. One recurring feature was Parekh’s creative excuses for why he was slow at getting things done. Arkadiy Telegin, cofounder of Leaping AI, wondered why it took him so long to do pull requests. Parekh replied that a military drone had damaged his house during fighting between India and Pakistan. Later, it turned out he was not near the conflict zone:
借口很常见。一个经常出现的功能是帕雷克（Parekh）的创造性借口，说明为什么他擅长完成事情。跳跃AI的联合创始人Arkadiy Pelegin想知道为什么他花了这么长时间才能进行拉力。帕雷克回答说，在印度和巴基斯坦之间的战斗中，一架军事无人机损坏了他的房屋。后来，事实证明他不在冲突地区：
Why the work was slow, allegedly. Source: Arkadiy Telegin on X
据称为什么工作很慢。资料来源：X上的Arkadiy Pelegin
It’s alleged Soham misled at least one employer about his location. A founder said they thought he was US-based, and even showed a company laptop sent to a US address. The laptop was later returned – Soham claimed it was sent to his sister. However, he never had a US address, and could not have one legally since he did not have any USA visa.
据称，索姆（Soham）至少有一个雇主误导了他的位置。一位创始人说，他们认为他是美国的，甚至展示了一家送往美国地址的公司笔记本电脑。笔记本电脑后来退还了 - Soham声称它被送给了他的姐姐。但是，他从来没有美国的讲话，并且由于没有美国签证，因此无法合法地讲话。
Igor Zalutski, founder of Digger (AI that understands infrastructure) gave me examples of what it was like working with Soham:
Digger的创始人Igor Zalutski（了解基础设施的AI）为我提供了与Soham合作的感觉：
“He said he applied for o1 visa and paid for premium processing so he was"few weeks at most" away from getting it. He said that his lawyers told him it's just better to be out of the US while the application is ongoing, and after it's done he's back, he"got an apartment in North Beach (San Francisco)". Later, a founder I background checked with told me that he [Soham] was telling them the same for over a year. And funnily enough, Soham started his dirty tricks from the very first day: he asked to push his onboarding meeting to the next day. He wrote how “this is unusual but it’s been unusually windy and rainy so I think that might have messed with the internet wire, I will have someone look over it tomorrow.”
“他说他申请了O1签证并支付了高级处理费用，因此他“最多只有几个星期”才能得到它。他说，他的律师告诉他，在申请正在进行中，他回到美国了，在完成申请后，他回来了，他回来了，他“在北部海滩（San Francisco）（San Francisco）（San Francisco）上了一年，我告诉了我的一年。索姆（Soham）从第一天开始就开始了他的肮脏技巧：他要求将他的入职会议推到第二天。
Digger AI terminated Soham’s contract a day or two into his onboarding. By that time, they had placed an order for an Apple laptop to be shipped to Parekh’s home address in India. Luckily, they could still cancel the order.
Digger AI终止了Soham的合同一两天，他的入职。到那时，他们已下令将苹果笔记本电脑运送到帕雷克在印度的家居地址。幸运的是，他们仍然可以取消订单。
4. Skyrocketing demand for AI engineers
4。对AI工程师的需求飞涨
One puzzling part of the story is how he got hired by Silicon Valley startups, particularly by “cream of the crop” places like Y Combinator AI startups. These businesses are well-funded, offer good base salaries close to the $200K/year range for senior engineers, and also attract more candidates than average. But the reality is complicated, and also explains why some parts of the tech market are still “candidate’s markets”, compared to the norm.
故事的一个令人困惑的部分是他如何被硅谷初创公司雇用，尤其是由Y Combinator AI初创公司等“庄稼奶油”的地方。这些企业资金充足，为高级工程师提供了良好的基本工资，接近$ 200k/年的范围，并且吸引了比平均水平更多的候选人。但是现实很复杂，也解释了为什么技术市场的某些部分仍然是“候选人市场”。
This is a reminder that becoming an “AI engineer” as a software engineer is straightforward enough: we covered Seven examples of software engineers picking up AI engineering, The “AI engineering stack” with Chip Huyen, and How Janvi Kalra became an AI Engineer at Coda — which later helped her get hired at OpenAI.
这提醒您，成为软件工程师的“ AI工程师”足够简单：我们介绍了七个软件工程师拾取AI工程师的示例，与Chip Huyen一起拿起AI工程学，“ AI Engineering堆栈”，Janvi Kalra如何成为Coda的AI工程师 - 后来帮助她聘请了Openai。
This is why the market is so hot for AI engineers:
这就是为什么AI工程师如此热的市场：
