Darshil patel retweetledi
Darshil patel
281 posts

Darshil patel
@Darshil2710
Member of Technical Staff @k_dense_ai
USA Katılım Temmuz 2011
173 Takip Edilen43 Takipçiler
Darshil patel retweetledi
Darshil patel retweetledi
Darshil patel retweetledi

We audited 67 public K-Dense Web use cases.
Question: what does K-Dense Web + human save versus human alone?
Answer: $3.08M of estimated analyst work delivered for $5,670 in credits. Median case: 560x lower cost, 107x faster.
Full audit: k-dense.ai/blog/k-dense-w…
English
Darshil patel retweetledi

Big research ambitions, tight budgets. 💸
Our Research Grant Program provides your lab or research group with:
✔️ 800 monthly credits pooled across your team
✔️ Unlimited seats
✔️ Full access to K-Dense Web
Apply at: k-dense.ai/research-progr…
...and tag a PI, lab manager, or research director who needs to see this! 👇

English

Excited to be part of the K-Dense team leveraging @Google’s new Multimodal File Search to deliver more powerful, verifiable AI capabilities for scientific research.
K-Dense@k_dense_ai
We're excited to be using Multimodal File Search in K-Dense. Read the blog by @Google on how we are using it to power search: blog.google/innovation-and…
English
Darshil patel retweetledi

Now you can use @huggingface @huggingscience directly in K-Dense Web, K-Dense BYOK and Scientific Agent Skills! Give the new hugging-science skill a try! github.com/K-Dense-AI/sci…

English
Darshil patel retweetledi

We're excited to be using Multimodal File Search in K-Dense. Read the blog by @Google on how we are using it to power search: blog.google/innovation-and…
English
Darshil patel retweetledi

Researchers don’t spend most of their time actually researching; they spend it reading, trying to map what’s already been said before they can say anything new.
@TimothyKassis lived that loop for years and eventually asked a simple question: What if the research process itself is the bottleneck?
So he built @k_dense_ai, an AI co-scientist that takes prompts like “draft a PhD proposal,” “map the literature on X,” or “find the gaps in Y,” and turns them into structured, long-form research outputs.
It scans 250+ databases, works across 200+ scientific formats, synthesises findings, suggests hypotheses, and produces technical documents that would normally take weeks.
The ambition isn’t to replace scientists; it’s to make sure they never have to work alone again. If the hardest part of research is navigating what already exists, this flips the equation.
Read the full story on @FinancialXpress: financialexpress.com/life/technolog…
English
Darshil patel retweetledi

Science isn't text. It's z-stacks, Western blots, LC-MS traces, lab-meeting recordings, whiteboard sketches, and voice memos from the PI. A text-only AI co-scientist has never actually been in the room.
Today @NVIDIAAI introduced Nemotron 3 Nano Omni, an open omni-modal model that fuses vision, audio, and language in a single reasoning stream. K-Dense collaborating to make to bring it to scientists worldwide.
How we're using it for agentic science:
k-dense.ai/blog/nvidia-ne…
English

Proud to be part of the @k_dense_ai team serving as NVIDIA’s evaluation partner for Nemotron 3 Nano Omni.
#AIForScience
K-Dense@k_dense_ai
A text-only AI co-scientist has never actually been in the room. Proud to be one of the partners evaluating Nemotron 3 Nano Omni at K-Dense, an open, omni-modal model with eyes and ears, deployable wherever your data lives. Super excited to be collaborating with @nvidia to bring Nemotron 3 models to scientists worldwide!
English
Darshil patel retweetledi

A text-only AI co-scientist has never actually been in the room. Proud to be one of the partners evaluating Nemotron 3 Nano Omni at K-Dense, an open, omni-modal model with eyes and ears, deployable wherever your data lives. Super excited to be collaborating with @nvidia to bring Nemotron 3 models to scientists worldwide!
NVIDIA@nvidia
English
Darshil patel retweetledi

K-Dense BYOK now has support for @NVIDIAAI Nemotron 3 Nano Omni.
Try it out: github.com/K-Dense-AI/k-d…
English
Darshil patel retweetledi

New in K-Dense BYOK: tabbed chats 🧪
Up to 10 parallel research conversations. Each tab, own model, history, cost. All tabs share the same project files. Streams keep running while you switch.
Open source, runs locally:
github.com/K-Dense-AI/k-d…
English
Darshil patel retweetledi

Introducing Pantheon: one question, 80 AI voices.
Ask a research question and Pantheon streams answers from diverse expert personas, surfaces citations, and shows where they converge or disagree.
We wrote up why we built it and how it works:
k-dense.ai/blog/introduci…
English
Darshil patel retweetledi

Darshil patel retweetledi

What if your agent can think like @demishassabis when thinking about AI for science? Take a look at the Demis mimeograph we created. github.com/K-Dense-AI/mim…
English
Darshil patel retweetledi

Frontier LLMs are smart. They're not anyone in particular.
Today we open-sourced:
• mimeo: reads the internet and writes a SKILL.md / AGENTS.md that clones how a person thinks
• mimeographs: 60+ ready-made experts (Jobs, Buffett, Wittgenstein, Regev…)
k-dense.ai/blog/introduci…
English
Darshil patel retweetledi

