Golden Grot Awards ceremony

And the Golden Grot goes to…. Our annual Golden Grot Awards celebrate the amazing dashboards our community creates. The winners in the two categories (personal and professional) receive their prizes and share the stories of why and how they built their dashboards.

Raj Dutt (00:00):

Hello again, and welcome to one of my favorite parts of GrafanaCON, the coveted Golden Grot Awards. These are becoming some of the most desirable awards and statues in the Observability industry. More popular and coveted than the Grammys, than the Emmys. This is serious stuff. And this is the fifth year that we're presenting these solid gold award. No, no, they're not really solid gold, but these gold statues that are designed to recognize excellence and innovation around our community. So let me grab the clicker since I didn't do that yet. And each year, the bar gets higher. And this year, it's clear that AI is shaping what's possible. So we've got four Golden Grot Award winners to present today, two dashboards and two projects. And so let's start this year's award in the personal dashboard category. So I'd like to congratulate Mohamed Adem, who has created something called the Aurora Chaser.

(01:03):

So I'm sure everyone here is familiar with the Aurora Borealis. Spectacular if you get a chance to see it. I went to Iceland a few years ago, and I did not get to see it in the entire four days I was there. But Mohamed has a solution for us all. Please welcome Mohamed Adem to present Aurora Chaser.

Mohamed Adem (01:25):

Appreciate it. Thank you. Thank you, everyone. My name's Mohamed Adem. I lead a software team at DRW. I am based out of Montreal, Canada, and I am this year's Golden Grot winner for the personal category. I'd like to thank everyone that voted. It's truly an honor to be doing this. So I started working with Grafana about three years ago, when I first interned at my company. I've used it a lot professionally, and this year a lot more personally as well. This year in specific, I wanted to build something to compete for the Golden Grot, so I built the Aurora Chaser. And if you haven't seen it already, it is a dashboard to track the northern lights. I think the northern lights are one of the coolest natural phenomenons we have, and I recommend anybody if given the chance to go out to see them. But as Raj said, it's very difficult to do so.

(02:10):

And throughout my research, I was able to find a lot of public indicators and websites around the scientific probability of the northern lights, but I wasn't able to find a composite view or indicator on whether or not these northern lights would be visible at X location. So that's what I decided to focus on. I decided to aggregate a bunch of public data to come up with my own landing page to answer the one question, "Should I go out tonight to see the Northern Lights?" It's probably a good idea to do a quick TL;DR on Northern Lights. I wanna preface this slide by saying that I'm not a scientist. This is what I understand so far. So yeah, in our solar system, we have this big giant star we call the sun, and the sun produces these charged particles we call solar wind. And we come to find out that solar wind has its own magnetic field.

(02:54):

And we actually have data around this magnetic field from our ACE and DSCOVR satellites parked at the L1 point. And we're specifically interested in the Bz component of this magnetic field. It basically tells us the north-south direction of the magnetic field. And what happens is, as solar wind approaches Earth's magnetic field, one of two things happen. If Bz is south or negative, then these particles get funneled into the poles, react to the gases in the atmosphere, and we get this cool light effect where oxygen would glow green, nitrogen would glow blue or purple, and we get the northern lights. If Bz is north or positive, well, nothing really happens. These particles get deflected, they don't under our atmosphere, and no northern lights. I live in Canada, so I was actually fortunate enough to catch the northern lights at some point in my life three years ago in Newfoundland, Eastern Canada.

(03:47):

This is a picture of me and my friends at like 3:00 AM in the morning, freezing to see the northern lights. But since then, living in Quebec, I have missed a few Northern Lights occurrences simply because I was asleep and I found out about it the next morning. So I do see myself using this dashboard in the future and hopefully setting up some alerts, to get paged when it's telling me to go outside. In terms of the setup, it's pretty straightforward. All my data is public. I mainly use NOAA or the North Oceanic Atlantic Administration. They publish data around the Bz component we spoke about, solar, wind, and so on. I use Open-Meteo for atmospheric weather data. So this is the cloud coverage, visibility, and so on. So when it's cloudy, you can't really see the northern lights. And I use NASA for some sun imagery.

(04:28):

I have a collection layer, running Telegraf, and I essentially pull these sources and write line protocol to InfluxDB cloud. And then in Grafana, I query all this data in SQL. And then for the sun imagery, I have a text panel essentially rendering it using a direct URL. So we can jump into the demo, please and thank you.

(04:51):

There it is. So there's a lot going on. With the time I have, I'll try to highlight the important things. I think the most important stat panel is this one on the top left. And essentially this is my own aurora go/no-go score. And essentially what I'm doing is, here, I'm querying the three variables, the K-index, which is the global aurora activity index; Bz, which is what we spoke about; and cloud coverage. And essentially coming up with a score from 0 to 100 to essentially then map them into statements, whether to stay home or drop everything and go outside.

(05:25):

Oops. And scrolling down, this is the sun imagery from NASA that I spoke about. I believe these get updated every 15 minutes. This one here is pretty cool, the magnetogram. If you see these dark spots here, these are actually solar flares. It's what produces the solar wind to start the reaction. Scrolling down, I have some more time series stuff. Here, I'm comparing the solar wind data from ACE and DSCOVR satellites. It's pretty consistent across the board. There's some discrepancies, but it's negligible to say the least. And all the way at the bottom, I have some live webcam panels that I found online for different cities. I actually got this idea from last year's winner, Ruben Fernandez. He had this in his dashboard. So thank you, Ruben. I don't know if he's here tonight. And all the way at the bottom, I have a table for all the data sources that I used and the links to those data sources.

(06:17):

There's one more panel I wanna highlight, probably this one here. This might not look familiar to you as a standard panel type in Grafana. What this is, is it's a Business Charts panel type. And if you're not familiar with Business Charts, it's a plugin developed by Volkov Labs, and now maintained by Grafana, I believe, and essentially allows you to import the Apache ECharts library and then configure any of their charts here in Grafana. If you're not familiar with Apache ECharts, it's essentially this JavaScript open source library that allows you to, and what Business Charts allows you to do essentially is write some JavaScript, configure these charts natively into Grafana. I believe there's a talk this morning, the people that were trying to do a Digital Twin of the Ocean, they said one of their limitations in their dashboarding was 3D or 4D graphs. There might be something here, honestly, so it's worth a look.

(07:07):

So if we can jump back to the slides.

(07:16):

Yeah, so key takeaways from doing this and building dashboards in general. I always like to start with a question. I think the most meaningful dashboards answer a specific question, so whether or not your server is down or should I go outside to see the northern lights; for collection, I think if you have public APIs, you can always point Telegraf to them and then you basically unlock free time series data. Nowadays, with AI-assisted development, you can generate Telegraf configs pretty easily. The Telegraf docs are pretty well written. And then for visualizations, whenever I hit a use case where I can't use a native panel type in Grafana, I reach out to Business Charts and Apache ECharts, and essentially you unlock 100 plus new visualizations. And that's pretty much it.

(08:01):

If you wanna access the dashboard, you can scan this QR code, it is public. If you have any questions, you can reach me out on my LinkedIn. I will also be at the Ask the Expert booth right after this, and thank you. Oops.

Raj Dutt (08:15):

Amazing stuff. I wish I had this when I was in Iceland. I wouldn't have been disappointed. But congratulations and amazing job. Give it up again for Mohamed.

(08:34):

Grab the clicker. Cheers, thanks. All right, can we get back to the slide deck, please? Okay, so our next Golden Grot winner is truly from a mission critical environment. A lot of times we say mission critical when we're talking about something that isn't really, maybe that mission critical, but this is literally landing something on the Moon that you don't get a do-over, you don't get to, you know, redeploy and try again three minutes later, right? The stakes are super high. So Jackson Sweeney, the winner of our next Golden Grot Award, built this Observability system for Firefly Aerospace's Blue Ghost lunar lander. And at the most critical moment, the actual landing, this dashboard became a primary decision support tool. And it really proved that observability and visualization, you know, wasn't just about monitoring systems, but actually enabling organizations to achieve like a actual outcome, right, and the whole world was watching this.

(09:41):

So I'm a huge fan of aerospace, and this really gets me excited. So please join me in welcoming Jackson Sweeney from Firefly Aerospace. There you go. Cool, yeah.

Jackson Sweeney (09:57):

I get walk-up music. Hey, everybody, my name is Jackson Sweeney. I'm with Firefly Aerospace. And today, I'm going to be talking to you all about the dashboard that enabled us to land on the Moon. At Firefly, I am a thermal engineer. So what I'll be talking about today, just a little bit about Firefly Aerospace, what do we do, and then talking a little bit more about the Blue Ghost lunar lander, and then I'll talk about the dashboard a little bit more, and then close out with some mission imagery, show you all some cool pictures that we got. So about Firefly Aerospace. We have three key pillars of our business. First being launch. We develop two rockets in-house called Alpha and Eclipse. Alpha is considered a small launch vehicle, and Eclipse is considered a medium launch vehicle. Both of these rockets, their primary goal is to deliver payloads, primarily satellites, into space.

(10:57):

Next, land, which is my area of expertise. And what we're gonna be talking a little bit about today, we developed the Blue Ghost lunar lander. It is the first commercial lunar lander to ever successfully touch down on the surface of the Moon. Lastly, we have orbit. We have a line of satellites that we call Elytra. These are satellites that can range anywhere from Earth orbit to lunar orbit and deep space for any kind of mission that it may need.

(11:26):

So about the Blue Ghost lunar lander. We design and build it all in-house at our facilities in Cedar Park, Texas, which is right outside of Austin. Blue Ghost Mission 1 is the only mission that we've done so far. It launched from Florida on January 15th, 2025, and successfully landed on the surface of the Moon on March 2nd, 2025, about six weeks later. Blue Ghost Mission 1 delivered over 100 kilograms of scientific instruments for NASA. There were some pretty cool things. There was like a drill that drilled into the surface of the Moon. There was a telescope that was looking back at Earth. We had a sort of vacuum-style device that actually sucked up some lunar regolith and did some science on it, and a few other cool things as well. At Firefly, we're planning for yearly missions to the Moon starting in 2027.

(12:20):

So next, we have Blue Ghost Mission 2 coming up, and 3, 4, and hopefully some more after that.

(12:28):

Here is a slide just talking a little bit more about the engineering behind Blue Ghost. I could talk about this for hours if they would let me, but I think all I'll say here is that most of everything that you're seeing on the screen we design and build and test all of that hardware in-house at our facilities in Texas. So the dashboard. This dashboard, like Raj mentioned, is incredibly critical to us. It was our tool that we used to verify the vehicle's health during the mission. So this allowed us, while we were in the operation center, to make informed decisions about the vehicle so we're not making risky decisions that could potentially end the mission. So in order to get all of this super important data in front of us, we, for this dashboard, which is thermal-focused, so what you're kinda looking at is mostly temperatures and also data about heaters that we have on board of the vehicle.

(13:29):

Queried over 500 different telemetry points. So like I mentioned, temperatures, heater info. Also wanna know about the power system of the vehicle, so how charged are our batteries, for example. And I'm also curious about how the vehicle is oriented in space, so what side of the vehicle is the sun shining on because that has a big effect on temperature. And another thing that's really important to us is being able to quickly access all of the data. So we're able to take all of these telemetry points. In using Grafana, we neatly organize them on just two screens.

(14:06):

Lastly, I have some pictures from our mission, starting here in the top left, that is when we launched into space. So you're kinda looking up at the lander right now. This is while we're still connected to the rocket. The rocket dropped us off into earth orbit, which is the picture that you're looking at on the top right. We're kinda looking down the side of the vehicle now. And then you could see, I believe that's the Pacific Ocean of Earth that we're looking at. Next, we jump over into lunar orbit where, again, you're looking at a kinda similar camera view where we're looking down the side of the vehicle. And here, you're actually looking at the far side of the Moon. And lastly, this picture here on the bottom right, this is directly after we landed on the Moon, probably taken maybe 10 minutes after.

(14:52):

And you can see the shadow of the vehicle. So the sun is behind us and our shadow is projected out onto the surface of the Moon. And that little dot that you see there in the background, that is Earth. And so with that, I'll say thank you. Super happy and honored to be here. Much thanks to the Grafana team and the Grafana community. Thank you.

Raj Dutt (15:20):

Excellent. Thank you, Jackson. Really inspiring stuff. And, you know, maybe you could add this to the next mission payload and we could get Grot on the Moon. I don't think NASA would approve, but good job. Thanks. Give it up for Jackson.

Jackson Sweeney (15:34):

Appreciate it.

Raj Dutt (15:35):

Yeah. All right, that was amazing stuff. So for our next two Golden Grot winners, this year's theme, of course maybe predictably, was using AI, right, to recognize the incredible ways that our community is using AI, leveraging AI in their workflows to create new capabilities. And we really wanted to extend the Golden Grot Awards to pick two really compelling projects and people. So our first new category is pioneering AI in Observability. And the next Golden Grot Award goes to Oren of TeleTracking, who's using Grafana Assistant to transform how their team triages, mitigates, and reports on incidents, cutting investigation time from days down to minutes. So please welcome Oren. Thank you.

Oren Lion (16:25):

Thank you, Raj.

(16:27):

Thank you, thank you. All right. Well, good afternoon, everyone. I'm thrilled to be here at GrafanaCON and very honored to be selected for the Golden Grot Award. When Grafana Assistant was released last year, I used it here and there. But now it's part of how we get things done every day to get acceleration on incidents. And to fully appreciate our need for speed, here's what we aim for at TeleTracking. We work to optimize the flow of patients through the healthcare system. So think of the logistics to enable patient care. Patients are admitted, transferred, and transported. Bedside care is coordinated, patients get discharged, and bed cleaning is scheduled. We like to say our platform is like air traffic control for hospitals. The availability and performance of our platform has an impact on patient care. My name is Oren Lion, and I lead logistics engineering at TeleTracking.

(17:39):

So as we move to work on our new platform, we still need to support our current platform. We're transitioning engineers, and so we have more responsibilities on fewer people. And to do that, we're looking to save time to get more done, and we're doing that through Grafana Assistant. There are two areas that we're focused on. One is the incident response lifecycle and the other is directional insights. I just wanna back up one second. Okay. Yeah, sorry about that. I think these slides got out of order. Yeah, but so what I was saying is that we're moving engineers to work on our new platform and so the current engineers have more responsibilities. And in order to do that, we're using Grafana Assistant, and we all go through the incident response lifecycle. And it starts like this. For most of you, there's a change, there's an anomaly that's detected by Grafana, alerting,

(19:00):

the on-call is paged, they go through recovery and root cause analysis. Once that's completed, they look back into a post-incident report, incident reporting, but then they look forward in the mitigation phase where they work to avoid having the incident recur. So mitigating defects. And now I'd like to demo the incident response lifecycle in areas where we've seen acceleration.

(19:37):

So we'll switch to the demo. All right, so in this case there was an incident and it was picked up by Luke. He declares an incident in Grafana Incident, and he sees that alerts are firing, and he's getting pager stormed. It feels like everything's going down. In the alert, he's got a link to the dashboard. And so he starts investigating and sees that the alerts start to self-resolve. But, as we all know, things that self-resolve can always fall back. So let's take a look at the dashboard he was looking at. In this dashboard, we see the anomaly. And while he's investigating, we bring up Grafana Assistant here in the side panel and we start the assistant analysis.

(20:30):

So, as that's running, I'll switch to the output of the assistant on that day. And we see a couple interesting things. First, we see that assistant picked up that this could be a traffic spike or a broker issue. So two hypotheses. And we come down here, and as it's going through its reasoning, it's actually showing us what it found. And it found that all three brokers in Kafka took a knife drop, took a hit. So at this point, it's ruling out that it was a traffic spike and that it was probably a cluster-wide disruption. Over here, it concludes that it was probably a blip in AWS, and we buy that so we don't need to spend more time doing analysis. Now, a couple things to note here is that the grounding is showing up right here in the Grafana Assistant report. We don't need to go outside, like, when you use Google Gemini, there's a source and you can check the grounding.

(21:42):

This is all right here in the assistant output. This is a pretty good post-incident report and it's saving us time on that front, too. So lastly, after an incident, there's a mitigating defect phase, right? So we've had an incident, we've recovered it, we've gone through a post-incident report, it's already baked into the incident. And for mitigating defects, we can also use Grafana Assistant to create an alert. So in this case, we had a noisy service. It was bearing error messages, and so we wanted to create an alert on a deviation in errors. We can easily do that in Grafana Assistant. It creates the alert, We look over the alert, and we see that it would be kinda hard for a human to read that, so we can ask Grafana Assistant to simplify. And Grafana Assistant is able to simplify as well.

(22:56):

Now, this wouldn't be enough for us to actually check into code 'cause we do Alerting as Code, and we can import that into Grafana Assistant. So in Grafana Assistant, to continue adding the annotations that we require in alerts, it has a link to the alert editor. And in Grafana, there are many features. Sometimes you don't know if they're out there, an assistant helps you find them. So we'll let assistant bring up the alert editor, and we could add in the annotations that we would need to check in the code. So over here on the left, you see the alert editor, and we can check in things like a runbook or a dashboard link.

(23:44):

All right, and switching back to the slides. So as we pivot the business, we're looking for fewer people to support more with task-by-task acceleration from Grafana Assistant. But how are we managing spend? Quick trip down memory lane. Two years ago to ObsCON in New York, this chart helped me see which application was eating all the money for Observability, and we used adapter metrics to optimize cost. Two years later, accelerated by Grafana Assistant after about four hours of prompting and Grafana's cost attribution feature, we can see application spend as a percentage of total spend, target costs optimization by team, and track results over time. And it's real. So we can do a demo on this, too.

(24:54):

Right, so in this chart, on the right, I've got one of the prompts from Grafana Assistant, but we can see an application as a percentage of total cost. We can see things like cost drivers, and the team can take this and work to optimize cost over time. Now, everybody wants to know how to prioritize. And since we have the over/under budget panel over here, since we are over budget, this becomes a higher priority. So I can honestly say that Grafana Assistant is helping us spend less time and get more done across any role, individual contributors as well as managers. Thank you. Great.

Raj Dutt (25:44):

All right, amazing. Thank you, Oren. I'm really glad to see that you were able to improve efficiency, and thanks for your support the last few years. So enjoy your Golden Grot. Good luck getting that through airport security.

Oren Lion (25:55):

I'll try, thank you.

Raj Dutt (25:57):

All right.

Oren Lion (25:57):

Right.

Raj Dutt (25:58):

Bye. Over here. Okay, we've got one more around the AI innovation category. So I just met this winner backstage for the first time about 20 or 30 minutes ago, and he was wearing a really cool device around his neck. And this is a privacy-focused AI assistant that records all the audio that's happening while you're going about your day. And I think this example really exemplifies that by using Grafana Cloud and Grafana Assistant Investigations, there's a really cool way to kind of optimize complex speech-to-text and LLM pipelines, ensuring performance, quality, and cost stay on track as AI systems scale. It's kinda cool. It's an example of AI on the edge, I suppose. So please join me in welcoming Dhananjay Yadav to the stage, the winner of our next Golden Grot, thanks.

Dhananjay Yadav (27:03):

Me too.

Dhananjay Yadav (27:07):

Thank you. Hello, everyone. Super excited to meet you. I'm Dhananjay Yadav. I'm the CEO and co-founder of NeoSapien. What you are seeing over here is the world's first AI wearable, which understands all my conversations, creates individualized knowledge graph to build a deep context. Now, how this is done is through this. So what is NeoSapien? We are Neo 1. Our AI wearable device sits with you seamlessly to understand all your daily conversation, be it telephonic, online or offline. And then on top of it, there is conversational intelligence, which, at a single point of time, we have four different AI services which are being performed for your diarization, transcription, and summarization. The pipeline, which we have created, is patented as well. And on top of this, we wanted to always be privacy-first variable because, for me, the hardware devices, the cell phones which we are using, are never privacy-first. I wanted to build a privacy-first device.

(28:08):

So how do you do it? You basically make sure that your audio lives in the memory and then gets destroyed. And then there is no way of going back to it, just the raw transcript and the traces are the only thing which could be figured out. So it's like forensics, isn't it? This is where our pipeline comes into picture. So to device, every conversation gets ingested into your cloud via Kafka. And then we go through speech-to-text and LLM processing. Again, this is done by four different AI processes. Our orchestration layer, which is taking place every time. And this then sits on our personalized knowledge graph algorithm, which we have created ourself, wherein now every entity, every relationship gets differentiated on the basis of facts, figures, and entities.

(29:02):

So now, what are the challenges? So because audio couldn't ever be stored, only the traces and logs, the constraint gave us rise to figure out how to get it more optimized and understand what is breaking, what is not. And that's where, for us, three big pillars were. One is latency. So, for example, today I'm having this conversation, and at that point of time, for any user, a conversation time needs to be always standardized. For example, for any conversation, if you are getting it within 13 seconds, it needs to be 13 seconds. And if even for one percentage of the memory, if it hits two to three minutes, the user ability or the user actually gets, doesn't get proper understanding of the product or, just for him, the user experience of the products falls off. Second thing is we as humans talk a lot, right? So for us, discard rate is very critical because we want to understand what are those conversations which need to be stored versus what are those conversations which don't need to be stored.

(30:06):

Things like if you are talking anything, right? You are making some noises, you're talking to third person, things which are not meaningful, those shouldn't ever come to our model to optimize it, right? So for us, discard rate is very critical. Second thing is cost. In our wearable device, which is basically working 24/7 with you, the most critical layer is optimizing the cost, and that's where our . So all the pipeline which works together needs to make sure that the cost is always standardized at every point of time. And if there is a variance, then there should always be an alert to tell us that there is a variance and we need to figure it out. So how do we do it? I will take you one by one. First is on the discard ratio, right? So what you can see is we log shape and not the content.

(30:57):

So, for example, you can see on the left-hand side the conversations, the duration of conversations. And on the right-hand side, what you see is basically the discard rate. When we say our median is around 11.5%, that is the most optimized scenario. But when it reaches to about 60, 70%, we need to understand why did it happen. Because if at that point of time we are not able to understand, then that memory might never come to customer, that conversation might never come back to the customer.

(31:30):

Second part is the pricing. Like I had mentioned, because there are four different AI services working at a single point of time, we want to make sure that every part of our orchestration layer, which it follows, should be standardized. If there is any leakage, we need to understand quickly otherwise it's a leakage, which, in our case, can result into a continuous leakage. The last part, and one of the most important part, is the latency. So for us, for any user experience, it needs to be standardized. They need to get memory within 30 seconds. Even for one memory, if it crosses two minutes, that is a massive alert. And that's where a user experience can actually dive down significantly, and that is something which we need to guard against. And that's when we started using Grafana Assistant Investigations, which really, really helped us out. We are a small team.

(32:28):

We had only four engineers. And with them, we were able to run the complete assistant program on our complete pipeline to understand what are some of those key bottlenecks across the complete system. And what we got was something very interesting. So when we ran, one of the examples which we got was an issue with our BLE advertising failures. So here, we realized that there were certain issues with our BLE, which was actually impacting the complete user experience. And that's where we were able to figure it out, fix it, and validate it directly in our code base. What I believe is that any partnership where you get learning about what is happening is not good enough, but it's about how they can help you in fixing it. And that's what Grafana has done for us. And for such a small team, we have been able to scale significantly, so much so that we are among the highest-selling wearables in India and now launching soon globally.

(33:28):

And yeah, what did this make us learn? Three important things. When there are intense constraint, like in our case there was no audio which could be stored, it actually gave us more clarity to understand what needs to be done, how do we need to log it better. Second is cost observability. For us, one of the biggest mod was per user cost needs to be very optimized, and that is what Grafana helped. And third is privacy is not equal to blindness. If you think that even if we cannot actually store audio, but we can still understand significantly through duration, discard rates, and latency. And this is what we have built and, yeah, super excited to share our learning with you guys. Thanks a lot and feel free to reach out to me.

Raj Dutt (34:18):

Thanks, Dhananjay. Great stuff. Careful, this one's really heavy. No, just kidding. Great job and all the best in scaling up your hardware products. It's always cool to see new players in the space. Congratulations on the Golden Grot. All right, and that's it for the Golden Grot. Hope you enjoyed it.