Microsoft Build: Transforming the future of startups with Microsoft AI

Summary

安妮·珀尔，微软公司副总裁，在Build会议上展示了微软如何通过提供尖端AI工具和资源，助力初创企业转型。她介绍了微软Azure AI平台，分享了Goodcall、Gretel和Nixtla等初创公司如何利用这些工具快速开发产品。特别强调了微软为初创创始人提供的资源，包括Azure学分、个性化指导和AI模板，以简化AI开发流程，鼓励创业者利用这些资源快速验证和扩展他们的创意。

Takeaways

• 😀 Annie Pearl 是微软的公司副总裁，领导着初创企业、学生开发者关系和技能提升团队。
• 🔧 微软致力于通过提供尖端的 AI 工具和精选资源来帮助初创企业，以简化 AI 开发流程并促进创新。
• 🌟 微软 Azure AI 平台是一站式 AI 开发需求解决方案，为初创企业提供了从 ChatGPT-4.0 到最新的 Phi-3 模型等突破性 AI 工具。
• 📞 Goodcall 是一家利用微软 Azure AI 服务的初创公司，他们使用 Azure AI 的语音识别技术将电话通话转换为可操作的对话。
• 🤖 Azure Machine Learning 提供了更深层次的模型交互性，允许初创公司使用自定义 GPT 并实时监控和优化性能。
• 🔏 Gretel 使用 Azure Machine Learning 构建了使用合成数据集开发和测试 AI 模型的方法，无需泄露隐私。
• 📈 Nixtla 是一家使用 Azure AI 基础设施构建 TimeGen 模型的初创公司，该模型可以提供时间序列预测。
• 🛠️ Microsoft for Startup Founders Hub 提供了工具、技术和资源，以及对创始人的支持，帮助他们构建变革性的 AI 应用。
• 🚀 微软正在使初创公司更容易地将想法转化为 AI 应用，通过 Founders Hub 提供快速启动模板和 Azure 学分。
• 📚 Lahini Arunachalam 演示了如何使用 Founders Hub 快速启动和构建 AI 应用，展示了从注册到部署的整个过程。
• 💡 微软鼓励所有有想法的初创公司创始人使用 Founders Hub，并通过 Microsoft Learn 平台继续他们的学习之旅。

Q & A

Annie Pearl 在微软的职位是什么？

Annie Pearl 是微软的公司副总裁，负责领导初创企业、学生开发者关系和技能提升团队。

微软如何帮助初创企业转型？

微软通过提供尖端的人工智能工具和配套资源来帮助初创企业，以简化人工智能开发过程，推动创新。

Annie Pearl 之前的职业背景是怎样的？

Annie Pearl 在整个职业生涯中领导了多个成长阶段初创公司的产品团队，最近担任首席产品官。

Goodcall 是什么，它如何使用微软的 Azure AI 服务？

Goodcall 是一个针对小型企业的 AI 驱动的电话助手，它使用微软 Azure AI 的语音识别功能将电话通话转换为可操作的对话。

Goodcall 的创始人 Bob Summers 对使用微软 Azure 的体验有何评价？

Bob Summers 认为微软 Azure 在对话质量、延迟和可靠性方面表现优异，他赞扬了 Azure AI 栈的易用性和对编码人员的友好性。

Gretel 是如何使用 Azure Machine Learning 来保护隐私的？

Gretel 使用 Azure Machine Learning 构建自动扩展和强大的访问控制，以便在不泄露隐私的情况下使用合成数据集开发和测试 AI 模型。

Nixtla 是一家什么样的公司，它如何利用 Azure AI 基础设施？

Nixtla 是一家专注于时间序列研究和部署的公司，它使用 Azure AI 基础设施构建 TimeGen 模型，该模型允许用户上传数据并立即获得时间预测。

TimeGEN-1 是什么，它在 Azure AI 生态系统中扮演什么角色？

TimeGEN-1 是第一个在公共云中的基础时间序列模型，它使得预测和异常检测变得更加容易、快速和准确。

如何通过 Founders Hub 快速将想法转化为 AI 应用？

Founders Hub 提供了 Azure 积分、个人指导和 Microsoft AI 工具，包括快速启动模板，帮助初创企业快速构建、发布和迭代他们的 AI 应用。

初创企业如何利用 Founders Hub 获得持续的支持和资源？

初创企业可以通过 Founders Hub 访问 Azure 积分、GitHub Enterprise、LinkedIn 福利、M365 生产力工具等资源，并且可以预约与 Azure 顾问和专家的一对一时间。

Outlines

00:00

😀 微软AI助力初创企业转型

Annie Pearl作为微软公司副总裁，介绍了她领导的团队专注于初创企业、学生开发者关系以及技能培训。她强调微软对初创企业的支持，提供尖端AI工具和精选资源以简化AI开发流程，推动创新。Annie分享了自己的职业背景和团队使命，即从创意到规模化阶段为初创企业提供AI和云服务以及个性化支持。她还提到了微软为初创企业创始人提供的资源中心Founders Hub，以及稍后将详细介绍的内容。

05:03

🛠️ 微软Azure AI平台：初创企业的AI工具箱

Annie Pearl深入讨论了微软如何通过Azure AI平台为初创企业提供支持。Azure AI平台是一个全面的AI开发需求解决方案，提供了从Azure AI服务到Azure机器学习，再到Azure AI基础设施的多层次服务。她举例说明了Goodcall公司如何使用Azure AI服务快速推出产品，并通过Azure机器学习引入更多交互性。此外，还介绍了Gretel公司如何利用Azure机器学习保护隐私的同时开发和测试AI模型，以及Nixtla公司如何使用Azure AI基础设施构建TimeGen模型，实现时间序列预测。

10:05

🌟 TimeGEN-1：开创性的时间序列模型

Nixtla的联合创始人Azul Ramirez介绍了TimeGEN-1，这是首个在公共云中的时间序列基础模型。TimeGEN-1能够理解时间序列数据并提供准确的预测，这对于商业机构和系统来说至关重要，因为它们通常以数字数据的形式交流。Azul强调了TimeGEN-1相比于传统模型的优势，并展示了如何通过几行代码快速实现预测。此外，还提到了TimeGEN-1 for Excel的发布，这使得在Excel中进行时间序列预测变得简单快捷。

15:10

🚀 从创意到AI应用：微软为初创企业创始人提供的支持

Lahini Arunachalam展示了微软为初创企业创始人提供的资源中心Founders Hub。通过Founders Hub，初创企业可以获得Azure积分、个人指导以及包括快速启动模板在内的微软AI工具。Lahini通过一个示例，演示了如何使用Founders Hub快速构建AI应用，从注册、选择模板、部署基础设施到更新应用，展示了整个过程的简便性。她还介绍了如何利用Founders Hub中的AI助手、Azure顾问和其他资源来支持初创企业的成长。

20:12

🌱 微软对初创企业全方位支持的愿景

Annie Pearl在演讲的最后部分重申了她对帮助初创企业的热忱，并概述了微软如何通过提供尖端AI工具和精选资源来支持初创企业从创意到规模化的整个旅程。她鼓励观众使用Founders Hub和Microsoft Learn平台继续他们的学习之旅，并期待看到大家利用微软资源构建的创新解决方案。

Keywords

💡人工智能

人工智能（AI）是指由人造系统所表现出来的智能行为。在视频中，人工智能是主题的核心，微软公司通过提供先进的AI工具和资源，帮助初创公司开发创新应用。例如，Goodcall利用微软的AI服务将电话通话转换为可操作的对话。

💡微软

微软（Microsoft）是全球知名的科技公司，视频中提到微软通过其Azure AI平台为初创公司提供服务。微软的产品和服务贯穿整个视频，展示了其对初创公司生态系统的支持。

💡Azure AI平台

Azure AI平台是微软提供的一站式AI开发平台。视频中强调了Azure AI平台为初创公司提供从基础服务到高级模型的全部AI开发需求，如Azure AI服务、Azure机器学习等。

💡初创公司

初创公司通常指新成立的、处于早期发展阶段的企业。视频中，初创公司是微软AI工具和服务的主要受益者，微软通过各种资源和支持帮助这些公司快速成长和创新。

💡创始人中心

创始人中心（Founders Hub）是微软为初创公司创始人提供的一个平台，提供工具、技术支持和资源。视频中提到，通过创始人中心，初创公司可以获得Azure积分、个性化指导以及快速启动模板等。

💡机器学习

机器学习是AI的一个分支，它使计算机系统能够从数据中学习和改进。视频中，Azure机器学习被提及，作为帮助初创公司通过自定义GPTs等高级工具来增强产品交互性的平台。

💡时间序列模型

时间序列模型用于分析按时间顺序排列的数据点，预测未来趋势。Nixtla公司在视频中展示了如何使用Azure AI基础设施来构建和部署他们的时间序列模型TimeGEN-1。

💡合成数据集

合成数据集是人工生成的数据，用于在不泄露隐私的情况下开发和测试AI模型。Gretel公司在视频中使用Azure机器学习来构建合成数据集，以保护隐私同时开发AI模型。

💡AI开发过程

AI开发过程涉及从构思到实现AI应用的一系列步骤。视频中，微软展示了如何通过创始人中心和Azure AI平台简化这一过程，使初创公司能够快速将想法转化为AI应用。

💡模型目录

模型目录是Azure机器学习中的一个功能，它允许用户访问和使用预训练的AI模型。视频中提到，初创公司可以通过模型目录找到并使用专门的模型，如Nixtla的TimeGEN-1。

💡云服务

云服务是指通过互联网提供的各种服务，包括数据存储、处理能力和软件等。视频中，Azure云服务为初创公司提供了必要的基础设施和工具，以支持他们的AI应用开发和部署。

Highlights

Annie Pearl 作为微软公司副总裁，领导初创企业、学生开发者关系和技能团队，致力于帮助初创企业通过微软 AI 转型。

微软深度投资于初创企业，提供尖端 AI 工具和精选资源，简化 AI 开发流程，推动创新。

Annie Pearl 的团队专注于为初创企业提供 AI 和云服务，以及个性化支持，覆盖从创意到规模化的整个创业旅程。

介绍了微软为初创创始人提供的资源中心 Founders Hub，帮助创始人构建变革性的 AI 应用。

讨论了如何通过微软 Azure AI 平台为初创企业提供访问尖端 AI 工具的途径。

Goodcall 利用 Azure AI 服务将电话通话转换为可操作对话，快速推向市场。

Goodcall 创始人 Bob Summers 分享了使用微软 Azure 的体验和对初创企业的建议。

展示了 Azure Machine Learning 如何帮助 Goodcall 增加产品交互性，提升用户体验。

Gretel 使用 Azure Machine Learning 构建隐私保护的 AI 模型，无需泄露敏感数据。

Nixtla 使用 Azure AI 基础设施构建 TimeGen 模型，提供即时时间序列预测。

Nixtla 的 TimeGEN-1 作为微软 Azure AI 生态系统中的新模型服务发布。

TimeGEN-1 是首个公共云中的时间序列基础模型，填补了 AI 在时间序列数据理解上的空白。

Azul Ramirez 展示了如何使用 TimeGEN-1 进行时间序列预测，并发布了 TimeGEN-1 for Excel。

演示了如何快速使用 Founders Hub 将 AI 应用想法转化为实际应用。

Founders Hub 提供 Azure 积分、个人指导和 Microsoft AI 工具，帮助初创企业构建、发布和迭代想法。

Lahini Arunachalam 展示了如何使用 Founders Hub 中的快速启动模板来构建 AI 应用。

介绍了如何通过 Founders Hub 获取 Azure 顾问和专家网络的一对一帮助。

鼓励初创企业创始人使用 Founders Hub 和 Microsoft Learn 继续他们的学习之旅和构建产品。

Transcripts

>> Excellent.

Great to see all of you.

Thank you so much for being here.

My name is Annie Pearl and I am

a Corporate Vice President here at Microsoft.

I lead our startups students developer relations and skilling team

and just really excited to be here

to close out this three days of Build,

talking about how we're helping to transform

the future of startups with Microsoft AI.

Here at Microsoft, we are deeply

invested in startups and helping them by providing

them with access to cutting edge AI tools and pairing that with

curated resources to streamline

the AI development process and

really usher in a new era of innovation.

We're going to go deeper into both of

these topics with our time together today.

But first a little bit about me and my team.

I spent my entire career leading product teams most

recently as Chief Product Officer

at multiple growth stage startups.

I'm incredibly passionate about helping startups and helping

to provide them with the right tools, technology, resources,

and support to be able to build

transformational applications across industries

from cybersecurity to healthcare,

life sciences, retail, and beyond.

My team is dedicated to helping

provide startups with access to cutting edge

AI and Cloud offerings combined with personalized support

across the entire entrepreneurial journey from

ideation all the way to scale and beyond.

This really starts with the Microsoft

for Startup Founders Hub where we

provide founders with access to those tools, technology,

resources and support to be able to

build transformational AI applications and we're going to

spend some time later in this talk

sharing a little bit more detail about Founders Hub as well.

Let's go ahead and dig in.

We're going to cover two topics today.

First, we're going to talk about how we're helping to

provide startups with access to cutting edge AI tools.

You're going to hear directly from

several startups about how they're using

these tools to be able to

ideate, to pivot and grow their businesses.

Then second we're going to talk about how we're making it

even easier to go from idea to

AI application and get your product

into the hands of those who matter

most which is of course your customers.

Let's kick off and talk about how we're helping to provide

startups with access to cutting edge AI tools.

This really starts with the Microsoft Azure AI platform.

It's a one stop shop for all of your AI development needs.

We're really excited that we're giving startups

the same access to all of

the groundbreaking AI tools that you heard about

over the last couple of days from

ChatGPT-4.0 all the way

to the latest Phi-3 models

that we talked about over the last three days as well.

Starting at the top of our stack,

we have our Azure AI services.

These are really intended to help you get started building

quickly regardless of your model selection.

They're prebuilt for specific use cases

and they're easy to use as making an API call,

allowing you to get started building really, really quickly.

I want to start off by talking about one of the startups in

our program Goodcall who's using our Azure AI services.

Goodcall is an AI-powered phone assistant

primarily targeted at small businesses.

To get started, they began using

our Azure AI speech recognition to be able to turn phone calls

into actionable conversations with

speech to text NLP and text to speech.

Because this was as easy to use as making an API call,

they were able to get this product out to

market in less than a week.

Let's hear now from Bob Summers who's the founder of

Goodcall a little bit more about his experience.

>> Every day small businesses get tons of

phone calls and on average they miss 65 percent of those calls.

What happens with Goodcall is that it's able

to answer those calls when the business cannot answer.

I want to find out if you have something in stock.

Can I get a table reservation?

That's like 80 percent of these repetitive calls that come in.

Then beyond that there's long tail query,

which could be like do you have a corkage fee for

wine or do you cut hair like curly hair?

The reason we chose Microsoft Azure is we did a benchmarking

against all of the other providers in

the market looking for conversational quality,

latency, and reliability.

Benchmark after benchmark, Microsoft just scored incredibly high.

In fact, in most of those benchmarks up to twice as fast.

Microsoft was founded by coders for coders.

It's really prevalent when you work with the Azure AI stack.

Meaning when you go into the studio,

you start selecting models,

you get your API keys,

you're building before you even know it.

That's really exciting for my team.

It provides us efficiency and speed to

build a product that is of

high production quality for our customers.

My experience with Microsoft for

Startups has been really wonderful

going to the Founders Hub to find

all the various resources that we might need.

We've had to work with the engineering tech staff to solve

certain challenges and problems and they've been very

receptive and helpful as we work through those problems.

My number one piece of advice for

startup founders I know you're an engineer,

but please don't build code first.

Talk to a customer. Always go through the lens of customers

before you build because

otherwise you're going to build something nobody wants.

>> As you heard from Bob testing, experimentation,

gathering customer feedback was really important to Goodcall.

After they got the first version of their product out there,

they gathered feedback and they saw

an opportunity to bring more interactivity into their product.

They began working with the second layer of our stack

Azure Machine Learning to bring

interactivity into their product experience

leveraging custom GPTs and this really allowed them to create

a more human-like experience that they could

monitor in real time and optimize for performance.

Like Goodcall if you're looking to work directly with

the models themselves or you

want access to more advanced tooling,

you can use Azure Machine Learning to be able to

find specialized models as part of our model catalog,

to be able to refine your prompts using tools like prompt flow.

With our Responsible AI dashboard,

you can make sure that your model is sufficiently de-risked before

you push your product out to production and

get it into the hands of your customers.

Let's take a look at another startup

using Azure Machine Learning. This is Gretel.

Gretel uses synthetic data sets to develop

and test AI models without having to compromise privacy.

They used Azure Machine Learning to build

auto scaling and robust access controls for

managed data and they were even able to work with customer data

directly using Azure's secure collaboration zones.

Let's dive a little bit deeper into how Gretel is using Azure.

This is a diagram showing the integration of Gretel synthetic data

into a data pipeline using

Gretel connectors and workflows with Azure.

These workflows take actions like

reading from Azure, training the model,

writing to Cloud storage and they can be run either manually

or as part of more of a kind of automized process.

This transformed data it replaces

sensitive information like names and addresses with

synthetic values that have many of the same characteristics of

the original data which is ideal for

data utility but also for data privacy.

This diagram really shows kind of that ease of automating

synthetic data into your model using workflows with Azure.

At the bottom of the stack we have Azure AI infrastructure.

This is where we let you build using

the latest open source or state-of-the-art

GPT models with the security reliability performance

that you would expect out of Azure.

Let's talk about our third startup Nixtla.

Nixtla use Azure AI infrastructure to build

their TimeGen model which allows anyone

to upload their data to an endpoint deployed in Azure.

This really helps be able to give you access

automatically to a time forecast almost immediately.

We're really excited to welcome Nixtla

to the Microsoft model catalog which

gives anyone access to their time series model

directly from inside Azure AI Studio.

We're going to hear now from Max and Azul who are the

co-founders of Nixtla before we turn it over to Azul for a demo.

>> I'm Max Mergenthaler. I'm CEO and co-founder of Nixtla.

Nixtla means time in Aztec,

one of the original languages of Mexico.

That's where I was born and raised.

Nixtla is a time series research and deployment company.

What we are doing is bringing

the generative AI revolution into data.

>> Nixtla started as an open source company.

We developed a lot of different libraries which are really

efficient in terms of

computational complexity but also are really accurate.

>> Time series research is

the systematic approach to try to predict the future,

to try to quantify

the uncertainty about what's going to happen next.

It is used for trying to predict

such diverse phenomena as ocean tides,

financial markets, and supply chain optimization.

>> We use a lot of

different technologies in the Azure and Microsoft ecosystem.

In particular, we need to

deploy our model to be accessible for other people.

We used Azure ML to host the model in that way

any person can just upload their data

and then you can have the forecast almost immediately.

>> Microsoft for Startups has helped us a lot.

It has opened doors.

But we have also had a great support

regarding the technological needs that we have as a startup.

We firmly believe that Microsoft's strategy

regarding artificial intelligence is highly structured and solid.

>> With that, I'd love to welcome to the stage Azul Ramirez,

CTO and Co-Founder of Nixtla.

>> Thank you so much for having us.

I'm Azul, I'm CTO and Co-Founder of Nixtla.

I'm also a proud Mexican trans woman.

Today I'm going to speak about TimeGEN-1,

which is the first foundational time series model

in a public Cloud.

Satya in his keynote released

Nixtla's TimeGEN-1 as a new model as

a service in the Azure AI ecosystem.

We were very excited,

it was a huge milestone for us.

But why foundational time series models

are so groundbreaking in our field?

As all we know,

current models like GPT-4o and Gemini only understand text,

image, and video.

They are not capable of understanding time series data,

and this is very important because at the end of the day

business institutions and systems don't speak in text,

they speak in numerical data,

in particular, in time series data,

which is the DNA of the world and help predict everything,

from sales to inflation.

As with learning in our society recently,

the status quo is not so good,

and this also applies in the time series forecasting field,

because before TimeGEN-1,

practitioners had to train their own models like ARIMA,

Prophet, LightGBM, or LSTMs.

Also they had to build and maintain

complicated pipelines from data cleaning,

perform hyperparameter tuning,

perform cross-validation, and model selection. Pre-trained models,

foundational time series models are the solution to this problem.

That's why a year ago we released TimeGEN-1,

the first foundational model for time series.

These models are the next frontier in artificial intelligence,

because they are capable of understanding

time series data and deliver accurate forecast.

Today, we are very excited to release with Microsoft TimeGEN-1,

the first foundational time series model in the Cloud.

TimeGEN-1 is an optimized version

of TimeGPT-1 for the Azure Cloud.

We are doing for numerical data what

Mistral and OpenAI did for text.

In a nutshell, forecasting and anomaly

detection with TimeGEN-1 becomes much easier,

faster, and more accurate.

Here you can see a comparison against Google,

Amazon, Salesforce, ServiceNow.

TimeGEN-1, TimeGPT-1 by

Nixtla is the best model across different datasets.

We will be making publicly available

the benchmarks in the next week.

Enough talking, let's write some code.

In this demo, I will be using

the Nixtla [inaudible] for Python,

you just have to install it using pip install nixtla.

I will be coding in BIM,

but of course you can use any other editor you want,

for example, Visual Studio Code, or Emacs.

Let me open BIM.

I'm going to import some libraries and I'm going

to load some data. Let me do that.

Here, I'm just importing some libraries,

then I'm going to write some data.

In this demo, I will be using electricity dataset.

Finally, I'm just going to plot the data.

Let's run this.

As you can see, we have electricity prices

for five different markets in Europe.

Of course, if you are in the electricity market,

you want to predict the future prices.

For example, if you are producing electricity,

you want to determine the future forecast.

Let's see how we can get those forecasts.

Before TimeGEN, as I mentioned before,

you needed to do a complete pipeline,

let me show you.

You need to import a lot of different libraries,

you need to perform different steps,

you need to parallelize your workload,

and it involves a lot of lines of code.

Let's simplify that by using the NixtlaClient class,

which is the bridge between the data and Azure AI.

I'm just going to import

the NixtlaClient from our Nixtla library,

and I'm going to instantiate the class.

Here is where the magic happens.

I'm just going to go to the model catalog.

As you can see, TimeGEN-1 is already available,

you can see the models that we have available.

Here is TimeGEN-1.

In just with one click on deploy,

you can deploy TimeGEN-1 on your own infrastructure.

Here I have a pre-deployed model.

Let me copy the URL and also the API key.

We are already authenticated.

Just like that, we can start forecasting

using the forecast method of the NixtlaClient class.

In this case, I'm going to pass the historic DataFrame

and also the number of steps I want to predict,

in this case, one day ahead,

which corresponds to 24 hours.

Let's run this. In this moment,

we will be seeing the forecast generated by TimeGEN-1

with just two lines of code in less than five seconds.

Thank you so much. We are also very excited

because we are launching TimeGEN-1 for Excel.

Let's predict some sales data using Excel and TimeGEN-1.

Let's open Excel.

Here we have a dataset.

We have basically the months and

also the values of the time series.

Here you can select the horizon you want to predict.

Here, for example,

I'm going to predict two years because it is monthly data.

I'm just going to click on "Make Prediction",

and just like that, you have TimeGEN-1 forecast.

As you can see,

the model captures the seasonality and the trend in the data.

TimeGEN also can significantly enhance large language models.

You can give ChatGPT the ability to understand

time series data in the future to TimeGEN-1.

Let's see an example.

Here we are uploading the forecasts generated by TimeGEN-1.

We are saying to ChatGPT plot that

time series data and also

plot the forecast generated but by TimeGEN-1,

and as you can see, we have a similar plot as before.

But let's make the things a little bit more complicated.

We can start using what if scenarios.

Basically we want to know the impact of

an increase of five percent in the demand over the price.

This is very important because,

for example, if you are working with sales data,

you can estimate the impact of some promotions.

Basically let's ask ChatGPT using

the forecast of TimeGEN-1 what will

be the effect on the prices

of an increase of five percent in the demand.

As you can see here,

we can see the original forecast,

but also the forecast with that five percent increase.

We can ask also ChatGPT what is

the exact increase with that five percent.

ChatGPT will give us this DataFrame to analyze further,

and we can have this awesome conclusion.

The exact rise in price with

five percent increase in demand will increase in

average the price by 10 euros per megawatt hour.

Just like that, we are start

dreaming on a real multi-modality world.

My name is Azul. It was a pleasure being here.

Thank you so much, and happy forecasting.

>> Thank you, Azul, that was incredible.

Like any startup,

Nixtla started out as just an idea.

I know many of you in the audience might be startup founders or

aspiring startup founders wondering how you

too can build and validate your ideas quickly.

Well, the good news is we're making it easier

now more than ever to be able to turn

your idea into an AI application within a matter of minutes,

so we're going to shift now from talking about

how we're helping to provide startups with

access to cutting-edge AI tools and begin talking

about how we're helping to streamline the AI development process.

As I mentioned earlier, this really starts with

the Microsoft for Startup Founders Hub where we

give startups founders access

to all the latest and greatest foundational models,

up to $150,000 worth of Azure credits and

unlimited one-on-one technical guidance and business guidance

so we can help you build not only a sustainable product,

but also a sustainable business.

Over the past year, we've seen a 10x increase in

the number of startups building using Microsoft AI.

To show you how we're making it even easier for you to be able to

take an idea and spin up

an AI application within a matter of minutes,

I want to welcome to the stage Lahini.

>> Thanks, Annie. My name is Lahini Arunachalam and

I oversee product for the Microsoft for Startup Founders Hub.

Open to all, Founders Hub gives startups access to Azure credits,

personal guidance and Microsoft AI tools including

quick start templates to help them

build ship and iterate on their idea.

I'm excited to demonstrate just how easy it is for

startups to get started building

their AI apps using the Founders Hub.

Let's say I'm a startup founder and I have an idea to connect

restaurants with their suppliers to

help eliminate waste and cut costs.

The first thing I'm going to do is I'm going to go

ahead and sign up for the Founders Hub,

and I first authenticate with LinkedIn.

Then I'm going to provide some information about my idea.

After I complete this,

I'll submit my application.

Once I'm accepted in the Founders Hub,

I can go ahead and check out my personalized homepage.

I want to get right to building,

so I'm going to click "Try a template"

and this lands me in the build with AI experience.

This is my one-stop-shop to get started

with everything I need to build an AI-enabled app.

First, I'm going to redeem my Azure credits.

This allows me to get started building on Azure

for free as I'm building out my startup idea.

Then I'm going to request access to

Azure OpenAI Service to get me

the latest GPT models running on Azure.

Once this is all set up,

I can take a look and explore these quick start templates that are

available to me that are designed

specifically for start up use cases.

I'm going to use this one which is

launch GPT app with your data because I'm

building a chat bot that connects

restaurants with their suppliers to help forecast that demand.

I can take a look at the template details including

the programming language as well as the AI model that it

uses and I can view

additional documentation directly in the GitHub repo.

I can take a look at the Azure architecture

as well as the user experience.

And it looks like this utilizes Azure command line or AZD.

I'm going to go ahead and open up VS Code.

In VS Code, the first thing I'm going to

do is clone the GitHub repo,

this has all the template code.

I'm going to initialize AZD and go ahead and

authenticate into Azure where I just drop those free credits.

Now all I do to deploy is actually type in

AZD up and this is going to deploy

all of the infrastructure as code as well as

a code sample directly to my Azure subscription.

After a few minutes,

this generates an endpoint so I can go ahead and grab

that end point and check out my app.

When I go and see the app I have

a chat bot that's built with the GPT model.

That's pretty cool because I was able to

deploy this in just a few minutes.

I can tinker around with it.

But I need to update this so that it better matches my use case.

I'm going to go back to VS Code.

The first thing I'll do is I'm going to include

the restaurant inventory file that I want this chat bot to read.

I'm going to go ahead and make a couple of CSS updates just to

better reflect the look and feel of the app that I'm going for.

Once I make these simple updates,

I'm going to go ahead and save them and I will just type in

AZD up again and this will redeploy my application.

After a few minutes, I'm going to go back to that endpoint that I

had created before and refresh now you see of my updated app.

Let's try a custom question.

How many tomatoes do I need tomorrow?

It looks like 10 crates. This is great.

This is a working prototype,

and when I'm ready,

I can continue to iterate on it and I

can grab the link and actually share it with

early advisors or early customers to get

the valuable feedback I need as I'm building out my start up idea.

Now, I can always go back to the build with

AI experience to continue exploring

Azure AI Studio as I continue to

iterate and I want to take a look at the model catalog.

I can do so directly from here, so that I can refine

and continue to use AI in my app in ways that make sense.

I can always check out Microsoft

Learn for valuable resources I need

to continue building on Azure

as I continue building out my product.

While my app is in the hands of my customers,

I can come back to Founders Hub

and continue to build out my company.

I can utilize the AI assistant within

Founders Hub to help answer

questions about my experience with Microsoft.

How do I continue to get more credits so they

don't eat into my runway as I continue to build my app?

I can also check out

all the other great benefits I have

access to in addition to Azure.

This includes GitHub Enterprise,

LinkedIn benefits to help me on

my go-to-market strategy and M365 for productivity tools.

I know I can get stuck sometimes,

so I always have the option to book time one on one

with Azure advisors and experts from

our curated expert network to help me from anything how do I build

this thing on Azure to how do

I start thinking about my go-to-market plan.

Remember Founders Hub is available to anyone with just an idea.

We're really looking forward to seeing what everyone will build.

Thank you. Now back to you, Annie.

>> Thank you, Lahini. To

close it out you know we started off this presentation with me

talking about my passion for

helping startups and helping to provide them with

the right tools and technology resources

and support to really build

transformational solutions across the entire journey

from ideation all the way to scale and beyond.

We hope that you took away today

some good insights around how we're helping startups by

providing them with access to cutting

edge AI tools and how we're pairing that with

curated resources to really streamline

the AI development process as you just saw from Lahini.

For those of you in the audience who are startup

founders or maybe you're just tinkering with an idea.

We encourage you to use the QR code on the left to access Founders Hub.

Your learning journey doesn't have to stop here.

For those of you who want to learn more,

on the right QR code,

you can access start up

specific learning paths within

our Microsoft Learn platform as well.

As Lahini said, we can't wait to see

what you-all build with us. Thank you so much.