Short Essays | Midas Technologies

Locus of Control and Hiring

Thu, 10 Aug 2023 00:00:00 +0000

At Midas Technologies, amid an atmosphere brimming with innovation and productivity, there lies a foundational belief: the profound influence of an internal locus of control. This belief goes beyond buzzwords and stands at the heart of our hiring philosophy.

The locus of control concept, introduced by Julian B. Rotter in the 1950s, pertains to an individual’s perception of what influences life events. Some attribute their outcomes to external forces like fate or luck. In contrast, those with an internal locus of control firmly believe their actions, decisions, and determination shape their future.

So, why does this belief play a pivotal role in our hiring at Midas?

People with a dominant internal locus of control are inherently self-motivated. Their conviction that their actions matter drives them to initiate and achieve beyond what’s expected. Their resilience shines brightly, especially during challenges, as they seek solutions rather than placing blame on externalities.

Their commitment to personal growth, stemming from the idea that their own development can alter life trajectories, resonates with our ethos at Midas. Such individuals naturally prioritize responsibility, transparent communication, and teamwork.

Importantly, an internal locus of control isn’t set in stone; it can be cultivated. Our goal at Midas is not just to recognize this trait but to actively nurture it. Through challenges, constructive feedback, and fostering a reflective mindset, we aim to strengthen this empowering belief in our team.

Midas’s commitment to the internal locus of control goes beyond acknowledging the role of external forces. It’s a celebration of individual determination and the conviction that one’s actions truly matter. We’re not just looking for employees at Midas; we’re seeking visionaries ready to shape the future. If this resonates, you could be the change-maker we’re searching for.

Authenticity in Content Creation in the Age of LLMs

Mon, 07 Aug 2023 00:00:00 +0000

In recent years, technology has brought about a paradigm shift in the way content is created, distributed, and consumed. Central to this revolution is the rise of sophisticated AI-driven tools like Large Language Models, which offer the ability to generate human-like text across a multitude of domains. But with such advancements, there arise critical questions around the very essence of content creation: What does authenticity mean in an age where machines can emulate human expression? How can creators ensure their voice remains genuine, and more importantly, distinguishable from algorithmic outputs?

At the heart of the matter is a perceived dichotomy between authenticity and automation. Authenticity, in the context of content creation, revolves around originality, genuineness, and the distinct personal touch that only humans can bring. Every piece of content reflects the thoughts, emotions, biases, experiences, and culture of its creator. This personal signature is what often resonates with the audience, creating a bond based on shared values, emotions, or experiences.

On the other hand, tools like LLMs are the epitome of automation. Built upon vast datasets and intricate algorithms, they can generate content that’s coherent, informative, and eerily human-like. Yet, no matter how advanced, this content is essentially a reshuffling of existing ideas and phrases without genuine consciousness or emotion behind it.

However, it’s essential to understand that the line between human-created content and AI-generated material isn’t always clear-cut. Many content creators have begun to harness the power of AI to assist in their creative processes. For instance, a writer might use LLMs to brainstorm ideas, generate outlines, or even draft portions of their work. Does this collaboration diminish the work’s authenticity?

Not necessarily. Authenticity is not about rejecting technology but about ensuring that the essence and core of the content are genuinely reflective of the creator’s voice and intention. Just as a painter might use new tools or mediums to craft their masterpiece, a modern content creator can use AI to aid their process, so long as the result is true to their vision.

To ensure authenticity in the age of LLMs, creators must be intentional about their content’s purpose and the role AI plays in its formation. It’s about leveraging AI as a tool rather than a crutch, and always being mindful of the line between genuine creativity and automated regurgitation.

Furthermore, audiences today are becoming more discerning. They crave genuine connections and can often distinguish between content that’s been crafted with care and that which feels generic or machine-generated. Thus, even as AI becomes a more prominent tool in the creator’s arsenal, it’s those human touches – the unique perspectives, the subtle nuances, the emotional depth – that will continue to captivate and resonate with audiences.

The age of LLMs and similar AI-driven tools challenges our traditional notions of creativity and authenticity. Yet, instead of seeing this as a threat, it’s an opportunity. An opportunity to redefine what it means to be authentic, to merge the best of human creativity with the capabilities of technology, and to craft content that, even in an age of algorithms, feels deeply, unmistakably human.

Focus on Product

Mon, 08 Feb 2021 00:00:00 +0000

The following memo was originally written and published in order to be distributed internally. We have included it here in its unedited form.

One of the questions members of my team often ask during their performance reviews is, “how do I get better?” This is a very good question. Whether you are a developer, researcher, trader or something else entirely, your compensation is directly correlated with your performance. If you are a high performing team member, you not only get paid more money, but you also position yourself to take on more important responsibilities in the future, thereby increasing the number of ways in which you can add value to the team. The cycle is virtuous, and everyone on the team benefits.

My answer to this question is very similar to the motto at Y Combinator, an early-stage startup accelerator: “Make Something People Want.” There is a lot of substance for us here, so let’s try to break it down.

On the one hand, the “People” in this motto can refer to our trading counterparties. All else equal, our counterparties want to be able to execute large order volumes within tight bid-ask spreads during all kinds of market conditions. As market makers and quantitative traders, we should acknowledge that our value and the money we earn stems from the liquidity we provide, and we should therefore strive to perform this service well.

On the other hand, the “People” in this motto can refer to ourselves. Perhaps one of the key differences between tech-driven market making companies, like ourselves, and other technology companies is that we are almost always our own end-users. We are the end-users of the network infrastructure we build. We are the end-users of the software we write. We are also the end-users of the machine learning pipelines we research and release. In our team, high performing people make things that people on our team want.

Exactly what people on the team want is almost besides the point. The important point, and I cannot emphasize this enough, is that if you want to get better, you must be product-oriented. No matter what your role is on the team, there is some tangible piece of technology that you can work on building and improving.

Why is being product-oriented so important?

One reason is that you maximize the return on your time. It is very easy in the quantitative trading industry to waste time. Several trading companies with very smart, hard-working people have failed because they spent their time on things that either did not matter or did not materialize. If you instead focus on building products, you are less likely to fall down that rabbit hole. The products you build are your deliverables. Your deliverables chart the path to meeting business objectives.

A second reason is that you more easily combine your efforts with other team members. Everyone works together on a single code base, building on each others’ work, making things better and better. Even failed products inform the direction of future work, and, if everyone publishes their work, two people are less likely to try the same thing the same way, fail twice, and waste time and effort.

If earning money were a convex optimization problem, a team of product-oriented people working together would be like applying stochastic gradient descent with a large batch size: every iteration is more likely to take a step in the “right” direction.

So, if you want to know how to get better, simply focus on building products. This means:

If you are a developer, focus on building applications, test suites, and docs.
If you are a researcher, focus on building quant trading libraries, machine learning pipelines, and data analysis libraries.
If you are a trader, focus on building trade automation libraries, monitoring tools, and parameter optimization libraries.¹

If you are new to a particular role and you don’t know how to do any of the respective tasks above, then you should concentrate on learning how.² And, if you discover you are not yet skilled enough to make meaningful contributions, then you should concentrate on improving your skills until you are.

Notes:

These lists are not meant to be exhaustive. ↩︎
Everyone on the team will help you along the way! ↩︎

On Poker, Mahjong, and Campus Recruiting

Thu, 14 Jan 2021 00:00:00 +0000

If you are a talented computer programmer or engineer currently undergoing campus recruiting, you are probably considering entry-level positions at a lot of different companies. You are almost certainly also considering a lot of different career paths, as your first job out of college has a big influence on what you learn, whom you meet, and your access to future professional opportunities. So, with so much at stake, how should you proceed?

Before we discuss strategy, perhaps it might help to first recall two definitions from probability theory. Recall from your Stats 101 class the definition of a random variable, a variable whose values depend on random outcomes. Also recall the definition of expected value, the probability-weighted average of possible values of a given random variable. Don’t worry if you don’t remember all the details right away.

What’s important to note is that the key to winning games of chance, games whose outcomes are influenced by randomizing devices like dice, is by maximizing expected value. Take poker as an example. In order to make the most money in the long run, you must make decisions that are more likely to have positive payoffs and less likely to have negative payoffs. Simply put, you must squeeze as much money as you can out of your opponents when you think you are ahead and lose as little as possible when you think you are behind.

Mahjong is similar. For those of you who do not know the rules, mahjong is like gin rummy. On your turn, you draw one tile and discard one tile. You aim to make certain combinations of tiles, which have different point values. And, as in poker, you win by maximizing your expected value. That is to say, you keep the tiles with higher probabilities of making high value hands and discard the tiles with lower probabilities of making high value hands.

Even chess, a game with no hidden information, is like a game of chance. Why do you think grandmasters study their opponents’ games? They are looking for strengths and weaknesses, so that in competition they can make moves to enter into positions that they are likely to win.

As for campus recruiting, many soon-to-be graduates opt for a suboptimal strategy. They apply for as many entry-level positions as they can, attend each and every interview, and see what sticks. The problem is that they view campus recruiting as a numbers game, probabilistic in nature, and winnable if they cast a wide enough net. This haphazard strategy might work well if the goal is to land any job whatsoever. But you can do better than that.

If you were instead to view campus recruiting as a game of chance, then you have a clear prescription for success. Let expected value be $\mathbb{E} = \sum_{i=1}^N P_{i} * Q_{i}$ where $P_{i}$ is the probability opportunity $i$ becomes an offer and $Q_{i}$ is how good it would be if opportunity $i$ were to become an offer.¹ Then, all you need to do is to focus on opportunities that are more likely to become offers and opportunities that you believe are good.²

In other words, focus on companies that want to hire you. Focus on figuring out what you want and how these companies might help you achieve what you want. Do not focus on companies where you do not fit culturally. And definitely do not focus on opportunities just because they seem prestigious.³ Focus on maximizing your expected value.

Notes:

Perhaps this definition is bit reductive, but I think it might still be useful for many people. ↩︎
What qualifies as good is different for different people, and how to figure that out for yourself is probably worthy of a separate essay in itself. ↩︎
Prestige stems from when other people, not necessarily you, think something is good. ↩︎

Gamma-Theta Tradeoff and Compensation

Sat, 02 Jan 2021 00:00:00 +0000

Since people interested in working in the low-latency trading industry often apply to multiple opportunities, they often receive offers from multiple companies. If you are fortunate enough to find yourself in this situation, you now have a tough decision to make: which job offer do you take? Which company do you choose, and how do you decide?

There is a concept in mathematical finance, specifically options theory, that might help with this decision.

Options have a property called gamma, which is the second derivative of the option’s price with respect to the price of its underlying. We don’t need to get into exactly what that means here, but suffice to say that gamma is a good thing. It lets you make more money as you make money and lose less money as you lose money.

You have probably encountered gamma in your everyday life. When you buy car insurance, for instance, you are buying gamma. With this particular type of gamma, you pay less for repairs after an accident. Similarly, when you buy a Costco membership, you are buying gamma, since you pay less with a membership for a same quantity purchase than you otherwise would.

In fact, perhaps gamma can explain why, in many cases, the rich get richer. Some of these wealthy people own gamma – the fact that they already have wealth gives them financial opportunities unavailable to those without. Perhaps a similar line of reasoning can explain why the poor, so to speak, often remain poor.

Gamma is not free. In options theory, the price of gamma is theta, otherwise known as time decay. In our previous analogies, if gamma is your car insurance policy, then theta is your monthly payment. And, if gamma is your Costco membership, then theta is your annual fee.

Theta is worth mentioning, because typically the more gamma you get, the more theta you pay. This observation, also known as the gamma-theta tradeoff, is the options theory version of the “no free lunch” theorem. Indeed, at least in the case of car insurance, high-risk drivers tend to pay higher rates.

The gamma-theta tradeoff is relevant to compensation. People at large companies typically have little to no gamma, but are collecting decent theta – they have little to no opportunities to grow with the company, but have decent salaries. In contrast, people at small companies have a lot of gamma, but are collecting less theta – they have a lot of opportunities to grow with the company, but have lower salaries.

So, the next time you are deciding among job offers from multiple companies, all else equal, consider asking yourself: “What do I value more, gamma or theta?” The answer might surprise you.

Low-Latency Trading and Automobile Racing

Fri, 04 Dec 2020 00:00:00 +0000

One of the interview questions we at Midas Technologies always ask our applicants is why they are interested in low-latency trading. One of the reasons we do this is because we want to see how much effort you have put into your interview preparation. When you dedicate significant effort into your preparation, you’re telling us that you’re naturally hardworking, always prepared, and that you could be a good future member of the team.

Another reason we ask is because we want to see how well you understand the role. If you clearly know what you’re applying for and therefore have actively chosen to pursue this career path, this indicates that you are dedicated and passionate. You’re both more likely to be happy working with us for a long time, and we’re more likely to be happy working with you.

But perhaps the biggest reason we ask applicants their reasons for joining the low-latency trading industry is to assess how similar their motivations are to our own. Why we choose to do what we do is important: strong motivation leads to results, compelling our best people to put in the extra time to improve the things we have, add support for the needed things we do not have, and work efficiently as a team to drive returns. This culture is a core component of Midas and makes us who we are.

So, why do we do what we do? To more fully answer this question, it helps to draw an analogy.

Low-latency trading is like automobile racing. In both racing and trading, a fixed time-interval event determines rank. In racing, the fastest car, driver, and pit stop team achieves first place and is considered the best. In trading, the team that makes the most money between market open and close on any given day is considered the best. In both racing and trading, much of what goes into determining rank happens outside said event. In racing, a lot of research and engineering effort goes into developing the fastest car, and, in trading, a lot of research and engineering effort go into developing the most profitable trading system.

Let’s consider why racers race. To some extent, they must enjoy the very act of racing, driving at high speeds, taking the optimal path around a course, and avoiding collisions. Many low-latency traders, myself included, enjoy trading. Trading is a lot less manual than it used to be, but we can still tweak certain parameters that directly affect our profit and loss. When markets are volatile, particularly when much is at stake, trading can be as fun and exciting as racing.

Race-car drivers must also enjoy the work outside of racing – for instance developing and testing improved hardware components. We traders also enjoy the trading-related, non-trading activities, like model-building and strategy development.

But perhaps there is a subtle, more significant motivator, at least for race-car drivers by profession. In fact, perhaps it exists not only for professional race-car drivers, but more generally for athletes at the highest levels of any competitive sport. Perhaps the realization that those among the best can indeed count themselves as being among the best, that in essence few others have worked hard enough to cultivate the necessary skill and domain knowledge to compete effectively, is a powerful fuel in itself. That is, once people reach a certain level of performance, they enter a virtuous cycle and become more motivated to perform for the sake of performance.

We are like these people. A significant part of why we do what we do is because, as far as we know, we are among the best. And that knowledge drives us.

We hope that you are motivated and excited by the same. When considering why you are interested in low-latency trading, think: “I could be among the best for reasons x, y, and z, and I want to be the best at what I do.”