A Conversation with Manychat - EuroPython 2026 Platinum Sponsor

Would [music] you mind telling us what your company is and what you're doing at Europyon? >> I have a microphone. I hope you can hear me. Uh, hey, my name is Vlad. Um, yeah, Manyhat is u we call it chat marketing automation platform, but basically we help our customers to automate their social media. It means like you have Instagram, imagine you have 5k maybe more followers. You're influencer and you want to get some value from your audience. You want to increase engagement of your audience and I'm sure you saw some posts in Instagram. Hey guys, comment like book and I will send you my like first book. And yeah, this is where many chats step in and we automate this kind of conversation. you want to automate your stories, you want to automate your posts or you want to automate your conversation in the direct messages in any so kind of any social media. Yeah, thank you so much. >> Did anyone else want to jump in on this question? That was great by the way. >> I think it was great. That actually covers the first question. So um where does Python sit in many chat stack today and what kind of systems is it powering? Do you use other languages as well? >> Yeah, I can answer this one. Um I'm Sergey. I'm engineering manager at Manyhat. Uh I am part of the team that's integrating AI features into Manyhat. So and actually I'm I was part of the initial team that built initial Python service uh powering this AI capabilities to the manage that product. So well to give some context what we do we build AI agents uh that um we can well we use uh pantic AI agents and we wrap them with some context some knowledge about the account and and then we are able to answer like based on like a certain message paths and flows we are able to give LM answers to the users so that's where we realized that made sense to use Python so we use fast API service um to for our endpoints and then we have our cues and and then our workers that run those agents. So essentially that's where that's one of the places where Python sits. Uh it is great because we have access to all the tools and libraries that are um constantly evolved and built in Python. So it made a lot of sense to use it. And for instance also like for knowledge knowledge retrieval we use um rack and pg vector. So it's very smoothly to integrate with with Python and also for observability or for like one of the challenges we have uh is that we want to route um uh how to say uh LLM requests. So we need to integrate with multiple providers. So we use uh light LLM as our main library to route its traffic and and yeah like uh it just using Python in that part of the stack just unlocks all these capabilities. I'm going to ask a follow-up question. So, >> can I add something? >> Of course. Go ahead. >> Just to say that we we also use machine learning, classical machine learning. So, I am my name is Amin. I'm in the machine learning team. I was the first hire in machine learning and uh so we have we have AI meaning LLMs but we also use classical machine learning from NLP to yeah text processing to like the whole stack and everything is is Python open source. So, May I also step in here because uh we use Python not only for ML or AI engineering, we also use it for data processing and analytics. And for example, we have pretty huge uh data team. Uh they use Python a lot for the data pipelines. and me and my uh another teammate, we also kind of so we transfer all of these data and insights uh using our uh Python uh service to our users, actual users. >> So you said you're working with AI agents in production. This is obviously an evolving technology and it has a lot of challenges. So would you mind describing some of those challenges that you might have dealt with and how you're solving for those? >> Yeah, so um it is a big challenge for us because we have uh many paths of or messages automations that uh we want that uh an agent or LLM can answer [snorts] but at the same time we want to ensure that our system stays reliable and that all the different integrations we do um we still maintain certain level of quality. Um what happens with dealing with agents and nondeterminism is that suddenly you start building those complex pipelines where you want to validate stuff but things you you realize that uh it is very there's a lot of flakiness in your CI/CD pipelines because of the nondeterminism. So something we how we approach it is that um and on top of this just to add extra context like uh we uh not so we have multiple agents that might be using different configuration different model different providers. So this adds complexity on top. So we want to validate that things um from an infrastructure p infrastructure perspective things stay up and running. So we validate at the right time but when it comes to uh integrating new features probably at that time it doesn't make sense to get like a real answer from LLM. Instead what we do is that we use uh piest and we with we use cassettes and we mock responses of the agents. What this allows us is that we know that our pipelines um so at the time that we introduce new feature it's just a matter of we record the answer from the LM it stays in a cassette and then our pipelines we know that every time we'll give the same output unless we change some of the behavior of the feature and this turned out to be one of the greatest let's say in infrastructure decision we've made because it allows us to move really fast uh interate with agents but at the same time stay stay very reliable That is so interesting. Um I think that many people who are working with agents are facing very similar problems and it's really interesting to see a company that's successfully overcoming those issues and using traditional Python tooling in conjunction with LLMs. >> So you said you work with Python extensively. Are there other languages you work with and does this pose any challenges for you? Yeah. Uh we uh as a company 11 years old and at that time as the most famous language was PHP and we have built our platform on this beautiful language and I I don't think that PHP is bad. It's bad when you can't cook it normally. Uh the but reality is that nobody wants to write in PHP anymore. Not everybody h and we met the problem that it was one of our motivation when we are moving to the python because python ecosystem is growing you can see this beautiful conference and you can uh like enjoy this how often you can see the big PHP conference I don't know yeah that's why we're um try to deeply integrate and migrate our main uh processing code base and dependencies on python And we're doing it accurate because we are holding you can imagine we are holding 40% of automations that goes through meta. It's Facebook, Instagram, WhatsApp. It's huge amount of money or money messages and money also. And uh to show our seniority I want to say that we have downtime of our system on PHP and the Python only when meta have downtime. That's that's what I want to say. Yeah. >> Thank you. >> Um, so when you are trying to select data to ingest into the platform, are there certain accounts that are better than others and how do you make a decision about that? >> Yeah, we have um it was super hard for us to understand ideal customer profile. I would say we spent two or three years and we always switched between oh we want to do e-commerce oh we want to work within infrepneur I don't know but now we fully understand that we live in creator world especially with AI everyone create this video photo I even this weird photo on the banners when you have a five or no six fingers sorry I have five they have six >> and uh Yeah, ideal customer profile looks for us like it's Instagram account with minimum 5k followers who post minimum like two stories per week >> and we see on historical data that these accounts have minimum five conversation with their followers and they're growing their their engagement. That's why in many chat right now we are building one more product. It's many chat for brands and um we see that big brands like Apple, Uber etc. they want to work with smaller influencer because they're expert in their domain. It's five key like followers as I said and big brands many chat and the influencer and we connect them and help them get money and show their performance transparently for big brands. >> It's amazing that such a simple signal can be so predictive. Um I've certainly found things like that in my data science career that really surprise you. Um and it's quite rewarding when you manage to find that pattern. Thank you. >> So why did Many Chat decide to sponsor Europyonthon this year and what does Python mean to you individually? >> Okay, so I guess I can answer for me individually but also for like so for many chat in general. So as my colleagues have said uh just now we used to have everything in PHP and we've uh we've we're we're moving to to Python. So Python is is an important tool for the company. Um and then from from a personal like at least for me personally I so my background is mathematics. I was in research and then I got into programming and the main language that I know is Python and this is my fifth conference I think in Python. The other ones were mostly PI data because data science. Um but yeah for me personally I love the community. So everything is free. Um Python is not behind like some guard rails or like essentially um like a money wall. uh it's just like the community uh creates tools um the for the Python this weekend Saturday and Sunday. So I'll be I'll be staying the weekend uh to contribute to sprint. So those are adding to open source uh packages. So I'm excited about excited about that and uh yeah specific for manyhat I don't know if we have any any other things you covered dash >> I covered it. Yes. So we're moving from we've moved from PHP to to Python. So Python is great but also yeah Python Python in general is great because it's an awesome community. >> You also probably uh mentioned that of course besides the sponsoring the conference we are also organizing a lot of uh different knowledge meetups related to Python also collaborating with PI data. So for us it's important to create this safe space where people could just come and share their knowledge uh and just network with each other and grow together. So this is something that we really believe in. Uh so yeah definitely besides building the product we're also very passionate about building the community >> I think people can find us in also in at Pi Pi data Amsterdam and Pyon Spain in Barcelona so come see us in Barcelona >> amazing >> so I'd like to highlight something so for inance myself I started my career working in Java and suddenly well I stumbled upon Python I um I really like that it's very welcoming language like it's very easy clean to understand and I think that that's probably like removing that barrier makes it that a lot of people can get interested and I like in the last years I realized like um such a big community that is behind Python you can see it in this type of events and conferences I think that there is like some shared culture of what it means Python and you can see it in everyone in all the interactions you have with people everything is so collaborative. Um, people are willing to share and I think that that translate ultimately in how we use Python and what is the language. someone else. [laughter] So yeah, I will put uh five cents from myself about why I actually value Python a lot and uh for me mostly it's more about applicability of Python because for example me I started as a QA automation engineer I used by test a lot I use it for some DevOps approaches and practices then I switched to back end engineering now I'm backend with kind of more or less focused on data stuff so if you notice I jump from one area to another and everywhere it's Python. So yeah, applicability is just so broad and I I love it. So thank you so much for being a platinum sponsor. It means so much to us as a conference and sponsors like you mean that community efforts like this can happen. People can come together. We can have the sprints. people can continue to work on Python together to share the knowledge. And yeah, thank you so much. It's it's really important. And speaking of being part of the community, you may want to bring more of the community into your company. You've told me that you're actually looking to hire. So, do you want to tell us a bit more about that? Yeah, we are hiring we are hiring a Python senior engineer for our new brands product that I mentioned before where we bring value for a big companies and connect them with the small influencers and this is a new product for us under the big umbrella of big company Manyhat 400 people 11 years uh old and super profit I would say. Yeah. And uh we are looking for a senior Python developer. You can choose our two beautiful offices in Amsterdam or in Barcelona. I in Barcelona prefer Barcelona. Sorry. Uh and yeah, if you have experience, we are looking only for one main uh skill which will describe your seniority. We're looking for people who are bringing problems up and solve them, not for the people who are waiting when they get the problem. This is who we are looking for. Please apply. Go Barcelona. >> Yes. >> Yeah. For me actually it seems like it's a totally unique opportunity because on one side it gives you creative freedom of the Greenfield project but at the same time it's under the big umbrella of many charts. So it's give you all the resources that the big company can provide you. So I think this is totally unique and it's nice opportunity to consider. >> Let me plug my team as well. So I'm in the machine learning and data engineering team and uh so we have two parts machine learning and data engineering and the data engineering part uh uh we're hiring so we're hiring for a data senior data engineer uh platform engineer uh relocation to Barcelona I believe uh because most of the team is in Barcelona for this role. Uh, Barcelona is great. I re relocated from London to Barcelona a year ago and so it's great. So I recommend it. >> But I wouldn't say that you should be relocated only to Barcelona because Amsterdam is also great. [laughter] >> I've never been to Barcelona, but I've been to Amsterdam and it is great. So, and to be honest, having heard the challenges you're working on and how you're approaching them, I mean, if I didn't have a job I liked a lot, I'd be tempted to apply. So, uh, I think everyone in the audience should definitely think about it. >> Thank you so much. >> Yeah. And, um, just to close things out, we talked about a lot of things today. We've talked about like the really interesting work your company's working on. We've talked about like the products. We've talked about the problems. And we've talked about what Python means for you. So, is there one final thing that you want the Python community to take away about many chat as we end this interview? >> Yeah, definitely something that we want the audience to remember when they will think about many chat is that many chat it's human built and AI powered and this is actually something that we were really trying hard to show uh by our booth concept. So the idea was that every single day we were changing a bit and tiny little details details of our booth. So at first uh we just showed the curtains full of different Python codes and then on the second day and the third day we will start slowly removing removing and show actually behind people behind the code people that actually build the product people that actually establish the organization and for us this is something that's very important and um talking about this human touch and human factor for the third day we just understood that a lot of engineers who came here for the conference they visited a lot presentations and they probably must be so tired. So for this we just wanted to create uh a space where people could just come, chill, relax, uh get some smoothies, ice cream, enjoy the conversations, and just have fun. So yes, [laughter] thank you. >> I can vouch the space looks incredibly fun. I watched it building from across the way over the last three days. Anything else? Uh can I Elsa [laughter] I just remember one thing is that um sometimes yeah when people uh stop by at our booth and what they see usually it's some of the activations that we prepared like I don't know some quiz uh some prompt challenges but uh there is something that's left behind uh behind the curtains is that how each of us contributed to this specific project. We had plenty of calls. We had plenty of meetings. And for me, honestly, this is something I never experienced before in any of the companies. It's the way how engineers were involved. You can't even imagine how much efforts and um I don't know love and passion Amin for example put into creating the career throw for participants to make them so beautiful. So like with a little tiny details that was amazing. Alisa, one of our also Python engineers, she prepared very interesting challenge. that a lot of people could enjoy and yeah challenge themselves by playing with the prompts. So yeah, this is something that's left behind and this is something that I personally really appreciate about the company. So yeah, thank you. >> The final word or maybe not. We are here for a long time. If you want to know the future of the Python and your future, I mean we'll help you in our company. Please join and enjoy this time with us as we enjoy. I work seven and a half years in the company which means a lot and yeah I would say that average time is four and a half in the company which is pretty great. [snorts] Thank you for watching this and I will pass Mike if someone wants to say something. I'm >> just quickly saying go Python and go open source community. >> [laughter] [music]