Composable, Malleable & Lean Compute Infrastructure for the Future
Mehmet Ozan Kabak, CEO & Founder
Apache DataFusion • Belgrade Meetup • 27 September 2024
Apache DataFusion • Belgrade Meetup • 27 September 2024
About Synnada
2
Mehmet Ozan Kabak, Founder & CEO
10+ years experience in machine learning, stream processing, and big data -- academic background in dynamical systems and optimization algorithms
Sami Can Tandoğdu, Founder & COO
10+ years experience in enterprise software, data &ai, cyber security, scaled commercial teams for high growth software startups
Infrastructure Team
3 members
Product Team
4 members
ML Team
4 members
Founded
2022
Investors
Collective Spark, DayOne, 500 EE, Expeditions,
Team
Company
HQ
Austin, TX
Open Source
Apache DataFusion
Unified data processing
1 PMC + 3 contributors
Mithril → Coming soon
ML compiler
4 contributors
Raised
$4.3M
Apache DataFusion • Belgrade Meetup • 27 September 2024
We are in the Data Chasm�Building/maintaining data intensive apps is complex
3
Apache DataFusion • Belgrade Meetup • 27 September 2024
Apache DataFusion • Belgrade Meetup • 27 September 2024
Example: Creating a traffic forecasting app
4
Apache DataFusion • Belgrade Meetup • 27 September 2024
Apache DataFusion • Belgrade Meetup • 27 September 2024
Enter → Apache DataFusion!
5
⚛️
Modular & Extensible | …and Performant! | Rust based | Amazing community |
| | | |
Nearly 6k!
When we joined…
Thanks to DataFusion we can deconstruct the data chasm!
Apache DataFusion • Belgrade Meetup • 27 September 2024
Apache DataFusion • Belgrade Meetup • 27 September 2024
Apache DataFusion: Our work
⚛️
Apache DataFusion • Belgrade Meetup • 27 September 2024
Apache DataFusion • Belgrade Meetup • 27 September 2024
Aras: Unified data processing
7
⚛️
Copy and paste, change source
Your PosgreSQL query runs on Kafka!
Apache DataFusion • Belgrade Meetup • 27 September 2024
Apache DataFusion • Belgrade Meetup • 27 September 2024
Synnada Composable, Malleable & Lean Infrastructure
8
Apache DataFusion • Belgrade Meetup • 27 September 2024
Apache DataFusion • Belgrade Meetup • 27 September 2024
Mithril: Composable ML library
9
⛓️ Build ML models with different backends, infrastructure, optimizers 🐞 Acts as a compiler, helps you debug without dry-runs 🎼 Compose, mix-and-match different models 🧩 Enables UDF generation and model interoperability 🚦Aligns any ML model with DataFusion 📖 Built in-house, soon to be released open source! | |
🤖
Apache DataFusion • Belgrade Meetup • 27 September 2024
Apache DataFusion • Belgrade Meetup • 27 September 2024
Workbench: Programmable data blocks
10
🎹 Composable blocks for data applications 😎 Can be used via UI or through APIs ⛑️ Works locally – in the future on public and private clouds ☁️ Prototype data applications locally ➡️ OOTB: Fault tolerance, low latency, (WIP) distribution 🏹 Community edition will be released in Q4 | |
🧱
Start prototyping locally
Graduate to cloud
Distribution
Fault tolerance
AutoML
Workspaces
Collaboration
Access control
Observability
Apache DataFusion • Belgrade Meetup • 27 September 2024
Apache DataFusion • Belgrade Meetup • 27 September 2024
Road Ahead for DataFusion: Our thoughts
11
Apache DataFusion • Belgrade Meetup • 27 September 2024
Apache DataFusion • Belgrade Meetup • 27 September 2024
Team
12
ML Team
4 members → Growing to 6!
Product Team
4 members
Will be working on:
- Upstream DataFusion
- Spark 2.0
Data Infrastructure Team
3 members → Growing to 5!
Will be working on:
- Model training and inference infrastructure
Apache DataFusion • Belgrade Meetup • 27 September 2024
Thank you
Apache DataFusion • Belgrade Meetup • 27 September 2024