Debarshi

VP/Fellow Engineer @DigitalOcean (ex-AWS)

[DigitalOcean @digitalocean]

Introducing the Open Intelligence Summit. We're bringing together AI founders, builders, and researchers to discuss everything open: weights, data, agents, infra, and tooling.

The intelligence layer is the product now. Own it.

October 12-13. San Francisco. Register here. 🔗 https://t.co/Zfp6UR9AxU

[Debarshi @DebarshiRaha]

Closed out Ray Summit with a happy hour our team co-hosted with @nvidia, @inferact, and @NousResearch.

Gave a talk on how @digitalocean & @vllm_project topped the Artificial Analysis leaderboard for model performance, and later Balaji demoed our Inference Router live!

Thanks to everyone who stopped by.

[Meghan Grady @MegRaeGrady]

Looking forward to this event next week with @nvidia @inferact and @NousResearch. Keep your eyes peeled, more to come 😎

[Debarshi @DebarshiRaha]

Inference routing is having its moment! 🚀

@DigitalOcean inference router is now cache-aware!

You don’t want to lose your warm cache for repeated context by switching models mid-conversation — that means higher latency and higher cost.

Cache-aware routing keeps the context where it belongs: warm.

https://t.co/ZlndAMw4rD

[Debarshi @DebarshiRaha]

@digitalocean This is what we were waiting for

[Debarshi @DebarshiRaha]

Inference routing is having its moment! 🚀

• OpenRouter announced joining Stripe this week! (congrats @alexatallah and team)
• At the Agentic AI Summit @ UC Berkeley, I broke down how we built the @DigitalOcean / Plano open-source inference router.
• Enjoyed reading "The Semantic Routing Moment" https://t.co/uiChknVD11 by @XunzhuoLiu — great piece on how the industry is converging toward letting users express a preference (speed/cost/accuracy) rather than picking a model.

[Debarshi @DebarshiRaha]

Speaking at Ray Summit 2026 in SF!

@digitalocean + vLLM (with @inferact ) hit #1 output speed on DeepSeek, per Artificial Analysis — plus NVIDIA Blackwell Ultra.

This year's Summit co-locates the first vLLM Conference. Come say hi.

#RaySummit #vLLM

[Debarshi @DebarshiRaha]

Launched! Congrats to the team

[What's New on AWS (Unofficial) @awswhatsnew]

Amazon Personalize Now Generally Available

We are excited to announce the general availability of Amazon Personalize, a machine learning service that makes it easy for developers to create individualized recommendations for users of their applicati... https://t.co/EzRfkfVZxM

[mza @mza]

Nice form. #reinvent #awsdeepracer

[Debarshi @DebarshiRaha]

Launched!

[AWS Events @AWSEvents]

Customers asked us to share https://t.co/KHhpJq3k01’s recommendation expertise. The result: Amazon Personalize. #reInvent

[Debarshi @DebarshiRaha]

@seanodw @2xjesselee @Google No!!!!

[Amazon Web Services @awscloud]

Amazon Translate is now generally available! Whether it’s a few words or volumes of text, Translate provides fast, accurate & affordable translation that scales with your needs. https://t.co/V4JC4gPOKx

[Amazon Web Services @awscloud]

Introducing Inter-Region VPC Peering. Now you can peer VPCs across AWS Regions! #reInvent https://t.co/h1HPY2Ow5J

[AWS Events @AWSEvents]

Introducing AWS DeepLens, the worlds first wireless deep learning video camera for developers #reInvent

[Debarshi @DebarshiRaha]

From an article, “Collectively, the five FAANG stocks had a market capitalization of about $2.8 trillion at the end of last week. The entire value of the U.S. stock market was about $28 trillion”

[Debarshi @DebarshiRaha]

Launched! #AWS #VPC support for #elasticsearch https://t.co/IUoPvkaON9

[Amazon Web Services @awscloud]

Now easily set up secure connections between Amazon Elasticsearch Service & your VPC! https://t.co/dtAQAoiTcE

[Debarshi @DebarshiRaha]

@nickstenning Right.

[Amazon Web Services @awscloud]

Amazon Elasticsearch Service now supports Elasticsearch version 2.3! https://t.co/hl5DAMBydU

[Amazon Web Services @awscloud]

Now add more instances & store more data in your Amazon Elasticsearch domains. https://t.co/6mCGOD0Oom