Natalie Vais
@nvais.bsky.social
gp at spark capital ✨ databases, infra, dev tools
Reposted by Natalie Vais
We've been working on something exciting in the Arrow/DataFusion ecosystem, which finally shipped with yesterday's release of DataFusion. You can now use Run-End-Encoded arrays in group by clauses!
github.com/apache/dataf...
github.com/apache/dataf...
Support Aggregating by `RunArray`s · Issue #16011 · apache/datafusion
Is your feature request related to a problem or challenge? It's currently not possible to aggregate by RunArrays. Example code grouping by a `RunArray` use arrow::array::{Array, Int32Array, RunArra...
github.com
January 13, 2026 at 9:45 AM
We've been working on something exciting in the Arrow/DataFusion ecosystem, which finally shipped with yesterday's release of DataFusion. You can now use Run-End-Encoded arrays in group by clauses!
github.com/apache/dataf...
github.com/apache/dataf...
Reposted by Natalie Vais
What a year!
From $25M funding to Git-like bucket forks, our 2025 was all about powering modern AI workflows with fast, flexible object storage.
Catch the highlights! 👇
From $25M funding to Git-like bucket forks, our 2025 was all about powering modern AI workflows with fast, flexible object storage.
Catch the highlights! 👇
January 8, 2026 at 9:34 PM
What a year!
From $25M funding to Git-like bucket forks, our 2025 was all about powering modern AI workflows with fast, flexible object storage.
Catch the highlights! 👇
From $25M funding to Git-like bucket forks, our 2025 was all about powering modern AI workflows with fast, flexible object storage.
Catch the highlights! 👇
Reposted by Natalie Vais
My EuroRust talk has been uploaded to YT! I cover how we architected our Rust database at @polarsignals.com as a set of state machines to enable deterministic simulation testing and catch hard-to-reproduce bugs before they hit prod.
A new talk is out! 🙌 @asubiotto.com shows how deterministic simulation testing (DST) uses random seeds and state machines to explore countless execution paths—and reliably replay failures when things go wrong 🦀🎲
👉 youtu.be/V12hHoagv-E
👉 youtu.be/V12hHoagv-E
Random Seeds and State Machines - Alfonso Subiotto | EuroRust 2025
Random Seeds and State Machines: An Approach to Deterministic Simulation Testing - Alfonso Subiotto at EuroRust 2025
Have you spent sleepless nights thinking about that heisenbug in production that…
youtu.be
January 7, 2026 at 11:46 AM
My EuroRust talk has been uploaded to YT! I cover how we architected our Rust database at @polarsignals.com as a set of state machines to enable deterministic simulation testing and catch hard-to-reproduce bugs before they hit prod.
Reposted by Natalie Vais
What an awesome year at Polar Signals! Some highlights:
📈 >10x data scale
⚡️ Shipped always-on GPU profiling
🦀 Shipped our new Rust/DataFusion/Vortex DB
Feeling lucky to get to work with such an incredible team & customers like Vercel, Notion, Canonical, Turbopuffer and more!
2026, we're ready!
📈 >10x data scale
⚡️ Shipped always-on GPU profiling
🦀 Shipped our new Rust/DataFusion/Vortex DB
Feeling lucky to get to work with such an incredible team & customers like Vercel, Notion, Canonical, Turbopuffer and more!
2026, we're ready!
December 31, 2025 at 5:11 PM
What an awesome year at Polar Signals! Some highlights:
📈 >10x data scale
⚡️ Shipped always-on GPU profiling
🦀 Shipped our new Rust/DataFusion/Vortex DB
Feeling lucky to get to work with such an incredible team & customers like Vercel, Notion, Canonical, Turbopuffer and more!
2026, we're ready!
📈 >10x data scale
⚡️ Shipped always-on GPU profiling
🦀 Shipped our new Rust/DataFusion/Vortex DB
Feeling lucky to get to work with such an incredible team & customers like Vercel, Notion, Canonical, Turbopuffer and more!
2026, we're ready!
Reposted by Natalie Vais
A simple 50-line code change saves us 30% of CPU across our whole infrastructure. This is the magic of system-wide profiling with @polarsignals.com. Easy to spot, easy to fix.
performance[vortex-array]: don't call is_valid to count bytes in varbinview by asubiotto · Pull Request #5814 · vortex-data/vortex
I noticed that the better blocks compressor uses count_referenced_bytes which calls is_valid on each view and results in an expensive scalar_at call. This was 30% of our system-wide CPU usage over ...
github.com
December 23, 2025 at 10:48 AM
A simple 50-line code change saves us 30% of CPU across our whole infrastructure. This is the magic of system-wide profiling with @polarsignals.com. Easy to spot, easy to fix.
Reposted by Natalie Vais
🚀 We recently announced our NVIDIA CUDA always-on profiler. Check out this blog on how to try it out super quickly with your CUDA workload today!🧊
www.polarsignals.com/blog/posts/2...
www.polarsignals.com/blog/posts/2...
Profiling NVIDIA CUDA in Kubernetes
The easiest way to get started Profiling CUDA in Kubernetes
www.polarsignals.com
December 18, 2025 at 3:59 PM
🚀 We recently announced our NVIDIA CUDA always-on profiler. Check out this blog on how to try it out super quickly with your CUDA workload today!🧊
www.polarsignals.com/blog/posts/2...
www.polarsignals.com/blog/posts/2...
Reposted by Natalie Vais
Today we're open-sourcing Durable Streams: a persistent stream primitive and HTTP protocol for reliable, resumable, real-time data streaming to clients.
We built it inside Electric. Now we're standardizing it as a standalone protocol.
We built it inside Electric. Now we're standardizing it as a standalone protocol.
December 9, 2025 at 7:25 PM
Today we're open-sourcing Durable Streams: a persistent stream primitive and HTTP protocol for reliable, resumable, real-time data streaming to clients.
We built it inside Electric. Now we're standardizing it as a standalone protocol.
We built it inside Electric. Now we're standardizing it as a standalone protocol.
Reposted by Natalie Vais
We completed a major project to switch our storage file format from Parquet to Vortex 🌪️ resulting in 70% average query performance improvement across the board 🚀
Learn more about how rethinking interface-imposed limitations unlocked these gains in our latest blog post 👇
Learn more about how rethinking interface-imposed limitations unlocked these gains in our latest blog post 👇
November 25, 2025 at 3:05 PM
We completed a major project to switch our storage file format from Parquet to Vortex 🌪️ resulting in 70% average query performance improvement across the board 🚀
Learn more about how rethinking interface-imposed limitations unlocked these gains in our latest blog post 👇
Learn more about how rethinking interface-imposed limitations unlocked these gains in our latest blog post 👇
Reposted by Natalie Vais
Awesome intro to Sync Conf from @thruflo.com, asking the question where sync fits in this fast changing world of AI.
Sync, as it happens, is a key component to this bright future!
Sync, as it happens, is a key component to this bright future!
November 12, 2025 at 6:40 PM
Awesome intro to Sync Conf from @thruflo.com, asking the question where sync fits in this fast changing world of AI.
Sync, as it happens, is a key component to this bright future!
Sync, as it happens, is a key component to this bright future!
Reposted by Natalie Vais
"If you can't count correctly with contention at high speed, it's not OLTP. And if you're not doing serializable, if you're cutting corners on isolation, it's not ACID."
Full talk: youtu.be/yKgfk8lTQuE
Full talk: youtu.be/yKgfk8lTQuE
November 21, 2025 at 1:09 PM
"If you can't count correctly with contention at high speed, it's not OLTP. And if you're not doing serializable, if you're cutting corners on isolation, it's not ACID."
Full talk: youtu.be/yKgfk8lTQuE
Full talk: youtu.be/yKgfk8lTQuE
Reposted by Natalie Vais
Aaron Francis captured it perfectly on Database School: with Tigris, ‘the region is just Earth.’
No CDNs, no cross-region replication, no complex DR— Reliability is the default on Tigris.
Link to the full episode below 👇
No CDNs, no cross-region replication, no complex DR— Reliability is the default on Tigris.
Link to the full episode below 👇
November 20, 2025 at 8:40 PM
Aaron Francis captured it perfectly on Database School: with Tigris, ‘the region is just Earth.’
No CDNs, no cross-region replication, no complex DR— Reliability is the default on Tigris.
Link to the full episode below 👇
No CDNs, no cross-region replication, no complex DR— Reliability is the default on Tigris.
Link to the full episode below 👇
Reposted by Natalie Vais
Excited to announce the TigerBeetle 1000x World Tour!
13 Cities in 6 Days, December 1-6
From AMS, BLR and CPT to VIE—join us!
tigerbeetle.com/event/1000x
13 Cities in 6 Days, December 1-6
From AMS, BLR and CPT to VIE—join us!
tigerbeetle.com/event/1000x
November 10, 2025 at 12:00 PM
Excited to announce the TigerBeetle 1000x World Tour!
13 Cities in 6 Days, December 1-6
From AMS, BLR and CPT to VIE—join us!
tigerbeetle.com/event/1000x
13 Cities in 6 Days, December 1-6
From AMS, BLR and CPT to VIE—join us!
tigerbeetle.com/event/1000x
Reposted by Natalie Vais
October in TigerLand is out!
- TigerBeetle’s Write Ahead Log utilization is doubled!
- The first Table Mountain Database Management Seminar in Cape Town
- @synadia.bsky.social, TigerBeetle pledge $512K over two years to the Zig Software Foundation @ziglang.bsky.social
mailchi.mp/tigerbeetle/...
- TigerBeetle’s Write Ahead Log utilization is doubled!
- The first Table Mountain Database Management Seminar in Cape Town
- @synadia.bsky.social, TigerBeetle pledge $512K over two years to the Zig Software Foundation @ziglang.bsky.social
mailchi.mp/tigerbeetle/...
November 7, 2025 at 12:58 PM
October in TigerLand is out!
- TigerBeetle’s Write Ahead Log utilization is doubled!
- The first Table Mountain Database Management Seminar in Cape Town
- @synadia.bsky.social, TigerBeetle pledge $512K over two years to the Zig Software Foundation @ziglang.bsky.social
mailchi.mp/tigerbeetle/...
- TigerBeetle’s Write Ahead Log utilization is doubled!
- The first Table Mountain Database Management Seminar in Cape Town
- @synadia.bsky.social, TigerBeetle pledge $512K over two years to the Zig Software Foundation @ziglang.bsky.social
mailchi.mp/tigerbeetle/...
Reposted by Natalie Vais
Ever wondered how your browser devtools jumps from bundle.min.js:1:27698 → src/index.ts:73:16? 🧊
We unpack the internals of JS Source Maps: Base64 VLQ encoding, mapping deltas, and how your original TS code gets reconstructed from minified bundles.
🔗 www.polarsignals.com/blog/posts/2...
We unpack the internals of JS Source Maps: Base64 VLQ encoding, mapping deltas, and how your original TS code gets reconstructed from minified bundles.
🔗 www.polarsignals.com/blog/posts/2...
The Inner Workings of JavaScript Source Maps
A deep dive into how JavaScript source maps work under the hood, with examples showing how all the pieces fit together.
www.polarsignals.com
November 4, 2025 at 3:07 PM
Ever wondered how your browser devtools jumps from bundle.min.js:1:27698 → src/index.ts:73:16? 🧊
We unpack the internals of JS Source Maps: Base64 VLQ encoding, mapping deltas, and how your original TS code gets reconstructed from minified bundles.
🔗 www.polarsignals.com/blog/posts/2...
We unpack the internals of JS Source Maps: Base64 VLQ encoding, mapping deltas, and how your original TS code gets reconstructed from minified bundles.
🔗 www.polarsignals.com/blog/posts/2...
Reposted by Natalie Vais
Peter Boncz's Vineyard Lecture:
- A brief history of the gains unlocked by generalization of compute
- The temptation towards specialization and a new era of GPUs
- What this means for DBMS design and embedded programming style
#TMDBMS
- A brief history of the gains unlocked by generalization of compute
- The temptation towards specialization and a new era of GPUs
- What this means for DBMS design and embedded programming style
#TMDBMS
November 4, 2025 at 12:52 PM
Peter Boncz's Vineyard Lecture:
- A brief history of the gains unlocked by generalization of compute
- The temptation towards specialization and a new era of GPUs
- What this means for DBMS design and embedded programming style
#TMDBMS
- A brief history of the gains unlocked by generalization of compute
- The temptation towards specialization and a new era of GPUs
- What this means for DBMS design and embedded programming style
#TMDBMS
Reposted by Natalie Vais
Philippe Bonnet's Poolside Lecture:
- NVMe SSDs today: high write amplification, spiky latency, weak guarantees
- Mismatched out-of-place internal writes and in-place block device interface
- The future: programmable, internals tuned for custom NVMe commands
#TMDBMS
- NVMe SSDs today: high write amplification, spiky latency, weak guarantees
- Mismatched out-of-place internal writes and in-place block device interface
- The future: programmable, internals tuned for custom NVMe commands
#TMDBMS
November 5, 2025 at 12:38 PM
Philippe Bonnet's Poolside Lecture:
- NVMe SSDs today: high write amplification, spiky latency, weak guarantees
- Mismatched out-of-place internal writes and in-place block device interface
- The future: programmable, internals tuned for custom NVMe commands
#TMDBMS
- NVMe SSDs today: high write amplification, spiky latency, weak guarantees
- Mismatched out-of-place internal writes and in-place block device interface
- The future: programmable, internals tuned for custom NVMe commands
#TMDBMS
Reposted by Natalie Vais
Tigris Data is proud to sponsor #RaySummit, where the next generation of AI is being built! Come meet the Tigris team and let’s talk object storage for agents, pipelines, and experimentation— the storage layer that scales with you!
📍 Marquis Marriott in SF today & tomorrow!
📍 Marquis Marriott in SF today & tomorrow!
November 4, 2025 at 8:06 PM
Tigris Data is proud to sponsor #RaySummit, where the next generation of AI is being built! Come meet the Tigris team and let’s talk object storage for agents, pipelines, and experimentation— the storage layer that scales with you!
📍 Marquis Marriott in SF today & tomorrow!
📍 Marquis Marriott in SF today & tomorrow!
Reposted by Natalie Vais
We built bucket forking so agents can safely use tools and run experiments without corrupting shared environments, especially when running in parallel.
Learn how every agent gets an instant, isolated sandbox to test, iterate, and scale without conflicts.
www.tigrisdata.com/blog/build-a...
Learn how every agent gets an instant, isolated sandbox to test, iterate, and scale without conflicts.
www.tigrisdata.com/blog/build-a...
Build better AI agents with bucket forking | Tigris Object Storage
Prevent rogue agent behavior by giving every agent an instant, isolated bucket fork—no data duplication, safe concurrency, and easy rollback with snapshots.
www.tigrisdata.com
October 23, 2025 at 3:59 PM
We built bucket forking so agents can safely use tools and run experiments without corrupting shared environments, especially when running in parallel.
Learn how every agent gets an instant, isolated sandbox to test, iterate, and scale without conflicts.
www.tigrisdata.com/blog/build-a...
Learn how every agent gets an instant, isolated sandbox to test, iterate, and scale without conflicts.
www.tigrisdata.com/blog/build-a...
Reposted by Natalie Vais
I had a great time talking about GPU profiling at Ubuntu Summit! Thanks everyone for joining!
October 24, 2025 at 10:58 AM
I had a great time talking about GPU profiling at Ubuntu Summit! Thanks everyone for joining!
Reposted by Natalie Vais
I could not be more excited about this. To our knowledge, the world's first open-source low-overhead always-on GPU profiler!
As always, zero-instrumentation and powered by #eBPF!
As always, zero-instrumentation and powered by #eBPF!
🚀 We're super excited to launch NVIDIA CUDA Profiling today! Check out the blog on how we made it happen! 🧊
www.polarsignals.com/blog/posts/2...
www.polarsignals.com/blog/posts/2...
Continuous NVIDIA CUDA Profiling In Production
Low-Overhead Profiling of GPUs with USDT Probes and eBPF
www.polarsignals.com
October 22, 2025 at 2:07 PM
I could not be more excited about this. To our knowledge, the world's first open-source low-overhead always-on GPU profiler!
As always, zero-instrumentation and powered by #eBPF!
As always, zero-instrumentation and powered by #eBPF!
Reposted by Natalie Vais
Will we see you at #PyTorchCon, Oct 22-23 in San Francisco?
Here are the five talks Tigris is most looking forward to, each one showcasing performance optimizations for AI workloads.
See all these and more, and come chat with Tigris about how we're building the storage layer for AI!
Here are the five talks Tigris is most looking forward to, each one showcasing performance optimizations for AI workloads.
See all these and more, and come chat with Tigris about how we're building the storage layer for AI!
October 14, 2025 at 5:57 PM
Will we see you at #PyTorchCon, Oct 22-23 in San Francisco?
Here are the five talks Tigris is most looking forward to, each one showcasing performance optimizations for AI workloads.
See all these and more, and come chat with Tigris about how we're building the storage layer for AI!
Here are the five talks Tigris is most looking forward to, each one showcasing performance optimizations for AI workloads.
See all these and more, and come chat with Tigris about how we're building the storage layer for AI!
Reposted by Natalie Vais
Big news! @tigrisdata.com raised $25M Series A led by Natalie Vais of Spark Capital w/ Martin Casado of a16z + Basis Set.
Tigris is a faster, freer, AI-native storage cloud– running on our own hardware.
Already powering fal.ai, krea.ai & Hedra.
Let’s build the open cloud for AI.
Read more on TC
Tigris is a faster, freer, AI-native storage cloud– running on our own hardware.
Already powering fal.ai, krea.ai & Hedra.
Let’s build the open cloud for AI.
Read more on TC
October 9, 2025 at 5:16 PM
Big news! @tigrisdata.com raised $25M Series A led by Natalie Vais of Spark Capital w/ Martin Casado of a16z + Basis Set.
Tigris is a faster, freer, AI-native storage cloud– running on our own hardware.
Already powering fal.ai, krea.ai & Hedra.
Let’s build the open cloud for AI.
Read more on TC
Tigris is a faster, freer, AI-native storage cloud– running on our own hardware.
Already powering fal.ai, krea.ai & Hedra.
Let’s build the open cloud for AI.
Read more on TC
Reposted by Natalie Vais
TigerBeetle at the New York Stock Exchange!
Last night, a panel and networking with @redpanda.com at 1792; this morning, a studio session with theCUBE.
Times we'll remember, together building the future of transaction processing.
Last night, a panel and networking with @redpanda.com at 1792; this morning, a studio session with theCUBE.
Times we'll remember, together building the future of transaction processing.
September 25, 2025 at 8:14 PM
TigerBeetle at the New York Stock Exchange!
Last night, a panel and networking with @redpanda.com at 1792; this morning, a studio session with theCUBE.
Times we'll remember, together building the future of transaction processing.
Last night, a panel and networking with @redpanda.com at 1792; this morning, a studio session with theCUBE.
Times we'll remember, together building the future of transaction processing.
Summer of OLTP continues...
- databricks acquires neon for $1B
- snowflake acquires @crunchydata.com for $250M
- mariaDB (re)acquires @skysql.bsky.social
- temporal acquires crystaldba
- vector (PE firm) buys @singlestore.com
- @supabase.com raising at $5B
turns out everyone needs txns 🙂
- databricks acquires neon for $1B
- snowflake acquires @crunchydata.com for $250M
- mariaDB (re)acquires @skysql.bsky.social
- temporal acquires crystaldba
- vector (PE firm) buys @singlestore.com
- @supabase.com raising at $5B
turns out everyone needs txns 🙂
September 11, 2025 at 4:24 AM
Summer of OLTP continues...
- databricks acquires neon for $1B
- snowflake acquires @crunchydata.com for $250M
- mariaDB (re)acquires @skysql.bsky.social
- temporal acquires crystaldba
- vector (PE firm) buys @singlestore.com
- @supabase.com raising at $5B
turns out everyone needs txns 🙂
- databricks acquires neon for $1B
- snowflake acquires @crunchydata.com for $250M
- mariaDB (re)acquires @skysql.bsky.social
- temporal acquires crystaldba
- vector (PE firm) buys @singlestore.com
- @supabase.com raising at $5B
turns out everyone needs txns 🙂
Reposted by Natalie Vais
Wrote up some thoughts for how Okta is getting squeezed from all sides: squished by Rippling at the bottom, Msft at the top, and other markets and startups in the middle.
I hope you love some red string diagrams 🤣
ventureinsecurity.net/p/the-unbund...
I hope you love some red string diagrams 🤣
ventureinsecurity.net/p/the-unbund...
The unbundling of Okta: are startups chipping away at Okta?
A guest post from Maya Kaczorowski who breaks down Okta’s competition and how Okta is not being unbundled, but rather squeezed from all sides.
ventureinsecurity.net
September 4, 2025 at 5:33 PM
Wrote up some thoughts for how Okta is getting squeezed from all sides: squished by Rippling at the bottom, Msft at the top, and other markets and startups in the middle.
I hope you love some red string diagrams 🤣
ventureinsecurity.net/p/the-unbund...
I hope you love some red string diagrams 🤣
ventureinsecurity.net/p/the-unbund...