Office Hours: I'm on the admissions team at Y Combinator and previously built machine learning systems at Lyft.Featured

Hi everyone! I’m Carolyn Conway, and I’m on the admissions team at Y Combinator, where I do everything from software engineering and data science to application review and interviewing founders.

I started out my career as a software engineer at YesGraph, which was backed by Y Combinator and then acquired by Lyft. I spent multiple years building machine learning systems at Lyft as a senior software engineer.

I graduated from Yale with a degree in linguistics, and later attended a bootcamp to break into tech.

Ask me anything about Y Combinator’s admissions process, non-traditional career paths, early stage startups, and more!

Thanks so much for joining us @coco!Elphas – please ask @coco your questions before Friday, July 16th. @coco may not have time to answer every questions, so emoji upvote your favorites πŸ”₯πŸ‘πŸΎβž•
Hi Carolyn! Great to meet you. I'd love to learn about YC's investment thesis around social impact startups. Is YC a good fit for social impact companies?
Hi Coco, Thanks so much for offering to answer questions! I'd love to hear your thoughts about characteristics of ML/DL startups that make them standout during admissions and in early-stage startups in general. - What types of ML products do you think are in need in the entire ML eco-system?- What are your top startups, or top features of startups, that solve traditionally non-tech problems using machine learning? I'd also love to hear your journey from doing ground-level data science to YC, e.g. what made you decide it's time to transition?
Hi Yi πŸ˜„I'm really hoping to see more products & technologies that make cutting edge ML/DL available to a wider array of companies. For companies like Open AI that are building state-of-the-art models, their biggest competitive advantage right now is that they have the funding to train bigger and bigger models. I think any product or technology that makes it cheaper to train large models will completely change the eco-system.
Hey @coco thanks for your time and answering our questions.I have an app idea where users can upload a photo of an outfit and the app tells them which designer created it, or suggests similar outfits. This is specifically for the South Asian wedding industry.Problem is - I'm not sure where to even start learning how to build something like that. Do I first have to figure out where to gather these photos? Any direction you can provide would be helpful.
@biancarods - as a South Asian, I am definitely intrigued to learn more about your application. Carolyn will also have a point of view but having worked in product, you need to start with the end-user in mind, and the pain-point you are trying to solve. From there you will be able to think through the high-level product features, translate them to user stories, and finally detailed requirements or tasks needed to build. Once you are at the requirement phase you can start to think through more details (i.e. -- where to store the photos, how to stores the photos, what types of formats would be needed), which would drive the final implementation of how to build. But start with the high-level workflow or experience of your user with the application, and continue to iterate through each of those levels. Feel free to DM me if you want to chat further!
@tsahai if you're interested, the current version of our application is a vendor search directory (think Angie's list), which can be found here:
Hey @tsahai thanks for those notes, super helpful. My question was geared more toward getting started with an ML-specific product and the initial steps involved to get a prototype up and running.
Hi Carolyn! I'm curious if you have any tips for how to best answer the "Please tell us about the time you most successfully hacked some (non-computer) system to your advantage" question on Y Combinator's application. I've met other women who struggle answering this because it almost sounds unethical (not saying that it actually is unethical, but it can bring up some negative connotations, 'hacking' a system to 'your advantage' definitely sounds like whoever was supposed to receive that advantage originally got the short end of the stick) but I've also heard that if founders leave this blank it's pretty much an automatic rejection. Maybe you could share some examples of how other people have answered this that made a good impression on you?
Hi Katya - I'm so glad you asked about this! We use the words "hack" and "advantage" extremely broadly here - this question is just trying to give you a chance to show that you aren't easily deterred by obstacles.We want to know that the founders we work with are creative and formidable problem solvers, so the best answers to this question describe a time when you were faced with a tough situation or unsolvable problem, and you found a creative way to solve it anyway.
Thank you for your response, Carolyn! It's SO helpful that you rephrased it like that, I was definitely focusing on "hack" and "advantage" too much and essentially misreading the question πŸ˜…
Hi Carolyn!Thank you so much for your time.In terms of applications that you have reviewed and founders that you have interviewed over the past 1 or 2 years, what trends have you observed in terms of startup ideas that took you a little bit by surprise?
Hi Nandana! One of the most surprising things about this job is seeing how different founders envision the future, and the last year has magnified that. I'm especially intrigued by startups who have found a need to address during the pandemic, and have a strong belief that the need will actually increase after the pandemic.
Hi @coco, thanks so much for doing this.I am the founder of a charity events startup called BrightOn. Is it preferable to have traction or proof of concept (i.e. through a waitlist) before applying to YC? Is it impossible to get in without either of those things?
YC definitely funds companies that don't have traction, so that alone shouldn't deter you from applying. However, I think a common failure mode for founders is trying to produce something perfect before showing it to users. Usually it takes very many iterations to build something people really love, so the faster you can get high-quality feedback from your users, the better! I'd suggest launching as soon as you have something that will be useful to even a handful of people.
Hi Coco, great to have someone here who can provide practical insight into how to get accepted by Y combinator. I would love some feedback on my application 101. Pls advise how to get this.
Hi Carolyn, I am a PM in clean tech and built a prototype for a climate tech startup that addresses critical goals like electrification and decarbonization. I haven’t launched officially, and don’t have users on my platform. I want to apply for YC this year, and need help in defining my strategy and finding a co-founder. Is it too early for me to apply? Should I apply when I have users and critical metrics?
Hi Sugandha,In general, there's no downside to applying to YC when you're very early on - we fund unlaunched companies every batch. It's also a big plus if you apply in the future and show lots of progress since the first application πŸ˜‰Whether you're in the batch or not, it sounds like Startup School could be really helpful for both defining your strategy and potentially finding a cofounder:
Thanks so much @coco! I’d love to ask about the non traditional career path. For front end engineers without a master or PhD in machine learning, what’s the best path to break in ML? For languages, should I focus on just python or also Java? Thank you!
Hi Ariell!In my experience, successfully breaking into a new domain requires (1) building the technical skills to do the work, and (2) finding people who are excited to take a chance on you and will think of you when new opportunities arise. I'll try to give my best advice on each of those:As for the technical skills, I'd recommend being as hands-on as you can in your learning. YouTube, Medium, and Kaggle are all your friends! I like to build my skills by working through a tutorial to build a basic model, and then trying to improve its performance iteratively, reading up on new concepts & testing them out as needed. has been a huge help for me.To the second point, I've found that new opportunities usually arise after I've provided some value with the skills I'm already known for. For years, I was the engineer building software around other people's ML models. Then when Lyft launched an internal ML program for engineers, the people I'd worked with on those projects already trusted me, knew I was interested in the domain, and had me top of mind to recommend for the program. There's no magic bullet, but I think you can set yourself up for good luck by finding work that's adjacent to where you want to end up.Lastly, no need to learn Java! Python will take you far 🐍
Hi Carolyn! Thanks for taking some questions. I have three quick ones about YC: Do you know if the next cohort of YC will be in-person? Is it considered best practice to have a YC alum vet your application before submitting it? What do you think it was about your application that helped you to be selected?We're building an automated financial planning experience for women. It will walk the user through building their plan and teach them about finance along the way, using their specific numbers. I was wondering if you could see machine learning applied to the platform we're trying to build? Currently, we do everything manually - we built the calculation engine, and we're working on automating parts of the process. My sense is we don't have enough data to 'teach' a machine. Best,Kristine
Hi Kristine! I've got 2 quick answers and one medium one πŸ˜‰The Winter 2022 batch will be our 4th remote batch.You don't need to have a YC alum review your application. The problem is that founders usually don't know *why* we chose to fund them, so some of their advice will be right and some will be wrong, and you'll have no way to distinguish between the two. However, a few years ago Dalton Caldwell gave a super actionable talk on how to write a great application, and I think it's still one of the best resources out there: YesGraph was accepted to YC in 2015, the founder had already taken another company through YC once before in 2008. So unfortunately I don't think the learnings there are very widely transferrable. πŸ˜”
Thanks for the great reply! Winter will be the last time we can apply then. As a mom of three little people, there is just no way I could do YC in person.
Hi Coco! I am the Mashia Tate, the founder of a career pathing/future of work tech start up called knoME. My burning question is this if you don’t have paying customers what other kinds of traction can be beneficial to mention on the slides?
- What are some qualities of applications that stand out from the rest?- How did you decide which bootcamp to attend? - What do you wish you knew before selecting a bootcamp?Thank you so much!
Hi Lirida,One of the best ways to stand out is to show that you can move fast. Even if I'm otherwise unconvinced by an application, I'm frequently inclined to interview founders who can make impressive progress relative to the time they've been working on the company.The deciding factor for me in choosing a bootcamp was finding the kind of environment that would be most effective for me personally to build the requisite skills. I'm a very auditory learner, so I tried to choose a bootcamp that involved more lectures and fewer readings, and one with a culture where I'd feel comfortable asking *lots* of questions.As for what I wish I knew - at the time I thought the only skill I needed was to get good at building software. In fact, I also needed to get good at technical interviewing. I wish I had known that those are very different skills. The bootcamp I attended has changed a lot since 2015, so I can't speak to how it is now, but at the time we spent a relatively small amount of time on learning to interview. That meant I had to do a lot of self-guided work online to build that skill, which is a much more difficult learning environment for me.
Thank you so much! your answers are so insightful!
Hi Coco! So happy to connect like this! I'd love to to get your thoughts on MVPs when it comes to attacking new markets. I'm a marketing professional who has worked for a few start-ups in the past. When we hit that need for growth, I've noticed teams will want to tackle potential markets with a stripped down product to test the waters so to speak. As a marketer, I struggle with this pace and messaging.What are your thoughts with MVPs when it comes to attacking new markets? And is this a usual practice of early stage startups at that point of growth?
I totally understand your discomfort around marketing a product that's perhaps not as built-out as you'd like. But usually my advice to startups is to power through the discomfort, and launch as soon as you have something that a user might get value from.One of the easiest ways for a startup to fail is by spending lots of time building a product that doesn't solve a burning problem for their customers. Given that, anything you can do in the early days to validate demand is essential. If your customers are still eager to use the product before it's fully polished, that's excellent signal that you're solving a real problem!
hi @coco thanks for contributing here! i'm working to build a startup in real estate that bridges knowledge workers to local, agrarian experiences to foster empathy between the burgeoning remote/knowledge class and traditional industries. kind of a vacation-meets-novel lifestyle integration-meets-incubator. i built my career in EdTech at a YC alum startup, worked on my own edtech startup (with the help of Startup School), and am now working at another edtech company that recently IPO'd. that being said, due to the nature of my startup being in tech and real estate, it's a challenge for me getting started, since i am not super technical and don't have an existing network in real estate/property. i'm doing my best learning, networking and drafting MVPs. but i am worried about the feasibility of this startup coming to life as a solo founder in a new vertical. it's my dream to realize this business with the help of YC. so my question is - how likely would it be for YC to accept a solo, non-technical founder based on this premise? what are some strategies for creating a stand-out application? any general advice for maximizing my chances? regardless of the odds, i will apply as many times as it takes, build my network, look for co-founders who complement my strengths/gaps, and continue to iterate and refine the concept. thank you again for sharing your knowledge and experience with us here :)
Is there a way to request you as a reviewer? I'm pretty sure every time we submit we're reviewed by people without very deep ML/DL software domain expertise.