5 Neat AI Startups From Y Combinator’s Summer of 2023 | TechCrunch

Image credits: TechCrunch

It’s that time of year again: the week that the startups in Y Combinator’s latest round show their products to the media — and investors — scrutinized. Over the next two days, a total of about 217 companies will be present, slightly smaller than the group of 235 companies last winter when VC enthusiasm cooled slightly.

In the first half of 2023, venture funds backed nearly 4,300 deals totaling $64.6 billion. That sounds like a lot. However, transaction value fell 49% compared to the first half of 2022 while transaction volume fell 35% year-on-year.

It’s worth noting that one segment – driven by hype and equal demand – is outperforming the others: AI.

According to CrunchBase, nearly a fifth of all global venture capital from August to July came from the AI ​​sector. And voracious appetite is on display in this summer’s Y Combinator team, with the number of AI companies doubling (57 vs 28) compared to the winter batch of 2022.

To understand what AI technologies are driving today’s investment, I dived into the summer of 2023, rounding up the YC-backed AI startups that to me are the most different – or the most numerous. most promising. .

AI Infrastructure Startup

Some of the startups in the Y Combinator W2023 group focus not on what AI can achieve but on the tools and infrastructure needed to do so. build AI from scratch.

For example, there’s Shadeform, which provides a platform that allows customers to access and deploy AI training and inference workloads to any cloud provider. Founded by data engineers and distributed systems architects Ed Goode, Ronald Ding and Zachary Warren, Shadeform aims to ensure AI jobs happen on time and at “optimal cost” “.

As Goode notes in a blog post on the Y Combinator website, the explosion in demand for hardware for developing AI models, especially GPUs, has led to capacity shortages. (Microsoft recently warned of service disruptions if it can’t get enough AI chips for its data centers.) Smaller vendors are online, but they don’t always provide resources. most predictable resources — making it difficult to scale across all of them.

Shadeform solves this problem by allowing customers to deploy AI work anywhere, on public cloud infrastructure. Leveraging this platform, companies can manage GPU instances across every vendor from a single pane of glass, configure “auto-reservation” when they have the machines they need, or deploy to a cluster of servers. with just one click.

Image credits: shade form

Another compelling Y Combinator startup tackling the challenges in AI operations is Cerelyze, founded by Sarang Zambare, a former Peloton AI engineer. Cerelyze is Zambare’s second YC after leading the AI ​​team at cashierless retail startup Caper.

Cerelyze takes AI research papers – the kind commonly found on open-access repositories like Arxiv.org – and translates the math contained therein into working code. Why is that helpful? Well, a lot of articles describe AI techniques using formulas but don’t provide links to the code used to put them into practice. Developers often have to reverse engineer the methods described in articles to build models and applications that work from them.

Cerelyze seeks to automate deployment through a combination of AI models that understand language and code, and a PDF parser “optimized for scientific content.” From the browser-based interface, users can upload research papers, ask Cerylize natural language questions about specific sections of the paper, generate or modify code, and run the resulting code in the browser. .

Now Cerelyze can’t translate everything in an article to code – at least not in the current state. Zambare admits that the platform’s translation feature currently only works for a “small set of papers,” and that Cerelyze can only extract and analyze equations and tables from papers, not numbers. But I still think it’s a fascinating concept, and I hope it will grow and improve over time — with the right investments.

AI Development Tools

Still developer-focused but not an AI infrastructure startup, Sweep automatically handles small developer tasks like high-level debugging and feature requests. The startup was launched this year by William Zeng and Kevin Lu, both veterans of the social network turning to the Roblox video game.

“As software engineers, we find ourselves moving from exciting technical challenges to mundane tasks like writing tests, documentation, and refactoring,” Zeng writes on the Y Combinator blog. “This is annoying because we know large language models [similar to OpenAI’s GPT-4] can handle this for us.”

Scans can detect code bugs or GitHub issues and plan to address it, Zeng and Lu say – writing and pushing code to GitHub via pull requests. It can also resolve comments made on pull requests from maintainers or code owners – a bit like GitHub Copilot but more autonomous.

“Scanning started when we realized some software engineering tasks were so simple that we could automate the entire change,” says Zeng. “Scan does this by writing the entire project requirements in code.”

Since AI has a tendency to make mistakes, I’m a bit skeptical of Sweep’s reliability in the long run. Luckily, neither does Zeng and Lu — Sweep doesn’t automatically implement code fixes by default, requiring humans to review and edit them before they’re pushed to the main codebase.

Applications AI

Transitioning away from the toolset subset of AI startups Y Combinator this year, we have Today, the company calls itself “the AI ​​co-pilot of corporate event planning”.

Anna Sun and Amy Yan co-founded the company in early 2023. Sun previously worked at Datadog, DoorDash and Amazon while Yan held various roles at Google, Meta and McKinsey.

Not many of us have ever had to plan a corporate event – certainly not this reporter. But Sun and Yan describe the challenge as arduous, exhausting and costly.

Sun wrote in a Y Combinator blog post: “Corporate event organizers are bombarded with countless calls and emails while planning events. “Pressed with tight schedules, planners are paying full-time assistants or tools for more than $100,000 a year.”

So, Sun and Yan thought, why not turn the hardest parts of the process over to AI?

Enter Today — provides details about upcoming events (e.g. dates and number of attendees) — can automatically contact venues and vendors and manage emails and calls relevant phone. Today it is even possible to take into account personal preferences around events, like amenities near a certain location and activities within walking distance.

I should note that it’s not entirely clear how it works behind the scenes of Today. Does AI actually answer, make phone calls, and respond to emails? Or are people involved somewhere in the process – to ensure quality, for example? Your guess is as good as mine.

However, Today is a very good idea with a huge potential market ($510.9 billion by 2030, according to Allied Market Research) and I’m curious to see where it goes next.

Image credits: Nowadays

Another startup trying to do away with traditional manual processes is FleetWorks, the brainchild of former Uber Freight product managers Paul Singer and Quang Tran, who previously worked on moonshot projects. at Airbnb.

FleetWorks targets freight brokers — the essential intermediaries between shippers and carriers. Designed to work in conjunction with a broker’s phone, email, and transportation management systems (TMS), FleetWorks can automatically book and track loads and schedule appointments with shipping facilities Lack of reservation portal.

Normally, brokers must contact by phone or email with drivers and dispatchers for shipments that are not automatically tracked to update on shipment status. At the same time, they must handle calls from trucking companies interested in reserving loads and negotiating prices, as well as scheduling appointments for unscheduled shipments.

Singer and Tran claim that FleetWorks can offload (no pun intended) by triggering calls and emails and transferring all TMS or email related information. In addition to sharing payload details, the platform can discuss pricing and book carriers, even call drivers and update the account team about issues.

“FleetWorks helps freight operators focus on high-value work by automating routine calls and emails,” Singer wrote in a post on Y Combinator. “Our AI-powered platform can leverage email or use human-like voices to make follow-up calls, handle workloads, and reschedule appointments.”

If it works as advertised then that really helps.