On October 18, 2025, we hosted the LLM 4 ETL Hackathon at the American University of Armenia. Participants in the hackathon had the chance to work with real data from a local brand. This gave them hands-on experience with data analysis and using LLMs for ETL and ELT. They did all of this using our MPP ETL tool.
We are a data engineering and AI/ML consulting firm based in Richmond, and our R&D office is in Yerevan, so it was exciting to bring this kind of event to our local community. It was Armenia’s first hackathon focused on LLMs and data engineering, and we were happy to have eLogiFest and Globbing as our sponsors.
Here’s a throwback to that day, to remind us how inspiring it was to see so many people learning, experimenting, and having fun at the same time.
Our team wanted this hackathon to be a different kind of event in Yerevan, so we started by bringing real data from a local company. We partnered with Globbing, a well-known e-commerce and logistics company in Armenia, and got anonymized data from their call center.
During the hackathon teams worked with these datasets using our MPP ETL platform with its LLM extension.
Our MPP ETL tool makes complex data pipelines simple and helps teams analyze data easily. It can collect data, clean it, and organize it so teams can use it for analysis. With the LLM plugin, participants could also use simple prompts to go through the data, ask questions, and find useful insights.
To add more excitement and a sense of competition, we gave every team the same token budget to work with, and displayed current budgets spent/left for all teams on a wide screen. Teams had to think carefully about how they used their tokens while building their solutions.
Participants could use different Large Language Models, ranging from best-in-class OpenAI to less expensive DeepSeek - 30 in total. But they had to be careful. Each team had a limited number of tokens, and the choice of model also affected the cost. Expensive models could use the budget faster, so teams had to ask questions carefully and avoid wasting tokens.
- Teams used LLMs to go through call center conversations to understand how customers felt, what they needed, and whether the customer issues were actually solved
- They pulled out the key points, and suggested ways to improve call-center operations
- They shared their thoughts about the tool and how it could be improved
This idea is now part of our MPP LLM Call Analytics, which helps call centers understand customer feelings and problems. If you want to see how it works behind the scenes, check out our blog.
The hackathon had a jury of IT experts to make sure judging was fair. Each team was scored based on clear criteria, and there was a scoreboard that helped keep everything organized. Judging happened in two steps of automated check and manual check.
The automated check was for giving quick feedback, but the jury’s review was the main scoring. Teams’ pipelines were tested with new data. This showed:
- How many tokens they used;
- Which models they picked;
- How many labels were in their ETL pipelines.
The manual review was the jury’s judgment based on five main things:
- Token use: how much each team spent and the models they picked;
- Implementation: how complete and well-structured the ETL pipelines were;
- Business value: how useful the solution was for solving data problems;
- Culture of work: Clarity and neatness of the work and whether it was completed on time;
- Overall approach: a mix of creativity and efficiency.
Teams could see their progress on the scoreboard. This made the event more fun and helped the jury focus on careful scoring.
Teams came up with lots of creative ideas and solutions during the hackathon. It was amazing to see participants learning new skills, trying new ideas, and working together in new ways.
Everyone left with something valuable from the day. Sponsors and supporters got a first look at future data analytics talent, and students gained practical experience, networking opportunities, and even chances for internships.
Participants also received a certificate of participation in the hackathon “LLM 4 ETL”.
Meet our Winners
- 1st Place: The Brainstorm team - Anastasiya Tikunova, Artur Abalov, Erik Shahbazyan, SARKIS TAMARYAN;
- 2nd Place: The AnalytIQ team - Vigen Mkrtchyan, Erik Nikoghosyan, David Khachatryan, Miqayel Hovsepyan;
- 3rd Place: The EXPM team - Sipan Muradyan, Armen Gabrielyan;
- We also had a special award from our partner e-LogiFest for the best value-for-money solution.
Meet Our Sponsors and Jury
We couldn’t have done this without our sponsors and expert jury. A big thanks to Globbing and eLogiFest for supporting the event.
Our jury helped guide the teams and made sure everything was fair:
- Harut Martirosyan (CTO at Firebird AI, Intent.ai)
- Mohamed Rashad (from Hyperion.ai)
- Stefan Papp (from Sophron Engineering)
- Karapet Elchibekyan (from eLogiFest)
- Sergei Shestakov (CEO of MPP Insights)
The Hackathon was a big step for showing Armenia’s growing strength in data engineering. We were the first company to host this kind of hackathon. We made it modern with our ETL tool, LLM extensions, and a smart judging system. Working with well-known brands like Globbing and eLogiFest made it even more special.
Who knows what other events like this we’ll have next? We’re always ready to be the bridge between technology and local talent.
