500 Global

24.5K posts

500 Global banner
500 Global

500 Global

@500GlobalVC

We invest in the world's potential. Momentum builds momentum. https://t.co/7mxfMs3E3d

SF Bay Area Katılım Temmuz 2010
3.1K Takip Edilen648.9K Takipçiler
500 Global
500 Global@500GlobalVC·
After bringing AI directly into the core of its platform by building their own products, @aircall is now deepening its voice AI capabilities through the acquisition of @vogentai. The move strengthens Aircall’s platform for more than 23,000 businesses globally, improving the reliability of high-volume customer interactions, enabling better qualification of inbound opportunities, and creating smoother handoffs between AI agents and human teams. The acquisition builds on Aircall’s growing momentum. Last year, the company announced it had surpassed $175M in ARR while accelerating growth to 25%, driven in part by adoption of its AI-powered products. This also reflects a broader momentum in AI: continued consolidation around specialized infrastructure players as companies look to deepen technical capabilities and strengthen real-world performance. Congratulations to @ScottChancellor and the Aircall team! Read the full story here: aircall.io/blog/news/airc… @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee
English
0
3
4
740
500 Global
500 Global@500GlobalVC·
Over the past year, Ricoh and @butlr_io have expanded their partnership into a broader effort to scale AI-powered workplace intelligence globally. Since Ricoh’s first investment in 2025, the companies have integrated Butlr’s privacy-first sensing technology into RICOH Spaces in Japan, enabling enterprises to better understand workplace utilization, occupancy, and collaboration patterns in real time. The partnership has also accelerated the adoption of physical AI infrastructure for more adaptive, efficient, and sustainable workplaces. Butlr Co-founder & CEO @deng_honghao shared, "Beyond data, we are evolving into a physical intelligence AI and infrastructure platform - turning real-world signals into automated actions that optimize comfort, energy, and operations. These insights impact every aspect of the office experience from cost and sustainability through collaboration, productivity, and comfort." Read the full story here: ricoh.com/release/2026/0… @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee
English
1
0
2
614
500 Global
500 Global@500GlobalVC·
This collaboration between @SakanaAILabs and @nvidia tackles a key inefficiency in transformer models: most hidden activations contribute little useful signal, yet GPUs still process them as dense computations. Instead of relying on traditional sparse methods that disrupt GPU execution, the team introduced TwELL — a sparse format designed to work with modern matrix multiplication pipelines — alongside custom CUDA kernels for training and inference. The result: over 20% speed improvements on H100 GPUs, with lower memory and energy usage. This category of infrastructure work matters deeply. As AI adoption accelerates, the winners won’t just be the companies building better models, but also the teams improving the economics of serving intelligence at global scale. @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee
Sakana AI@SakanaAILabs

How do we make LLMs faster and lighter? Don’t force the GPU to adapt to sparsity. Reshape the sparsity to fit the GPU! ⚡️ Excited to share our new #ICML2026 paper in collaboration with @NVIDIA: "Sparser, Faster, Lighter Transformer Language Models". This work introduces new open-source GPU kernels and data formats for faster inference and training of sparse transformer language models: Paper: arxiv.org/abs/2603.23198 Blog: pub.sakana.ai/sparser-faster… Code: github.com/SakanaAI/spars… While LLMs are undoubtedly powerful, they are increasingly expensive to train and deploy, with a large part of this cost coming from their feedforward layers. Yet, an interesting phenomenon occurs inside these layers: For any given token, only a small fraction of the hidden activations actually matter. The rest approximate zero, wasting computation. With ReLU and very mild L1 regularization, this sparsity can exceed 95% with little to no impact on downstream performance. So, can we leverage this sparsity to make LLMs faster? The challenge is hardware. Modern GPUs are optimized for dense matrix multiplications. Traditional sparse formats introduce irregular memory access and overheads that cancel out their theoretical savings for GEMM operations. Our contribution is twofold: 1/ We introduce TwELL (Tile-wise ELLPACK), a new sparse packing format designed to integrate directly in the same optimized tiled matmul kernels without disrupting execution. 2/ We develop custom CUDA kernels that fuse multiple sparse matmuls to maximize throughput and compress TwELL to a hybrid representation that minimizes activation sizes. We used our kernels to train and benchmark sparse LLMs at billion-parameter scales, demonstrating >20% speedups and even higher savings in peak memory and energy. This work will be presented at #ICML2026. Please check out our blog and technical paper for a deep dive!

English
0
0
4
2.6K
500 Global
500 Global@500GlobalVC·
FarMart reached an incredible milestone of $400M revenue run rate and EBITDA profitability in Q4 FY26, reflecting how sustainable growth can be achieved through technology and AI. By connecting smallholder farmers in India directly to agri buyers, @farMart_ helps address critical challenges in food security, supply chain resilience, and income disparities. Their impact is tangible: in 2025 alone, the company empowered over 4.8 million farmers, facilitated more than Rs 2,340 crore in payments, and helped prevent over 18,000 metric tonnes of food loss. FarMart is also scaling across the Global Majority regions such as Africa and the Middle East, and plans to expand further in the consumer products space through acquisitions. As @alekhs22, Co-founder & CEO of FarMart shared, "Our belief is that the future of agrifood will be built on AI-led coordination, disciplined execution, and capital-efficient infrastructure that can serve the ecosystem at scale." Congratulations to Alekh, @mehtab_hans, and the entire FarMart team on this important milestone. Learn more about how we're partnering with governments and investing in companies like FarMart to drive sustainable growth: 500.co/strategy/susta… Read the full story here: yourstory.com/2026/05/farmar… @CourtneyPowell @khailee @vishalharnal @almmura @atxfrench
English
1
1
4
1.1K
500 Global
500 Global@500GlobalVC·
For years, most biological models have treated DNA, RNA, and proteins as separate systems. But biology doesn’t operate in silos — sequence, structure, function, and cellular context are deeply interconnected. That’s what makes @PolymathicAI’s MIMIC compelling. MIMIC is a generative foundation model trained on LORE, a multimodal dataset aligning ~13M RNA transcripts and ~15.5M proteins across genomic, transcriptomic, proteomic, and semantic modalities from over 6,000 organisms. Instead of modeling one slice of biology at a time, MIMIC learns relationships across them, opening up new possibilities in protein engineering, splicing, RNA structure prediction, and drug discovery. @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee
Siavash Golkar@SiavashGolkar

1/12 Introducing MIMIC: a SOTA foundation model trained natively across DNA, RNA and proteins. MIMIC is multimodal and generative: it can use structure, regulation, evolution, and experimental context to infer missing biology or design new sequences.

English
0
2
12
2K
500 Global
500 Global@500GlobalVC·
In collaboration with @OpenAI, we hosted 35 AI-native and AI-enabled startups at 500 Global's Social House in Kuala Lumpur. The group included a mix of founders, builders and residents of 500 Global's AI residency — all exploring how frontier models can unlock new products, workflows, and entirely new categories of companies. Led by OpenAI's APAC Head of Startups Thomas Jeng, the session featured a deep dive into OpenAI’s latest models, a look at the company’s roadmap, and practical discussions around how startups in APAC can best collaborate and build alongside the platform. As AI capabilities continue to advance, one of the biggest opportunities ahead is helping more founders across the region access the tools, infrastructure, and networks needed to scale globally. Learn more about our AI residency here: 500.house @khailee @vishalharnal @AriefJohan @atxfrench
500 Global tweet media500 Global tweet media500 Global tweet media500 Global tweet media
English
1
0
10
1.1K
500 Global
500 Global@500GlobalVC·
Innovation is no longer concentrated in a few geographies. Companies like @trans_celestial are proof that frontier technologies in space, connectivity, and defence are being built and scaled from new global hubs like Singapore. Their expansion into the U.S. highlights how deep tech founders today are building globally — they are bridging ecosystems, talent, and mission-critical infrastructure across borders. We’re excited to see Transcelestial deepen their presence in the U.S. and what @jharohit and team build next! @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee
transcelestial@trans_celestial

The US government is looking to Singapore for deep tech. @ChannelNewsAsia featured us as one of the companies they're working with tapping capabilities in space & defence that don't exist in many places in the world. That's why we're opening in Austin. 🇺🇸

English
2
4
13
2.8K
500 Global
500 Global@500GlobalVC·
Big unlock for developers building with open-weight models. Making high-performance inference accessible directly within existing workflows is exactly how we accelerate the path from research to real-world deployment. The combination of interoperability, minimal friction, and transparent pricing is what will continue to push the AI ecosystem forward. @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee
DeepInfra@DeepInfra

DeepInfra × Hugging Face DeepInfra is live on @HuggingFace Inference Providers. Run DeepSeek V4, Kimi-K2.6, GLM-5.1 and 100+ more open models straight from the Hub — same OpenAI-compatible API, same low per-token pricing, no markup. Just add :deepinfra to the model name.

English
1
2
3
1.6K
500 Global
500 Global@500GlobalVC·
As AI reshapes healthcare, one of its most meaningful impacts is breaking down barriers to access — ensuring the right patients can find the right treatments at the right time. @MassiveBio's partnership with @OpenAI is a powerful step in that direction. By transforming complex clinical trial protocols into machine-readable formats, the team is enabling real-time, automated patient pre-screening at scale. "By collaborating with OpenAI, we are transforming how trials are interpreted and operationalized. This is infrastructure-level deployment that can close the gap between the thousands of open trials and the millions of patients who never learn they may qualify." — Co-founder & Chief Medical AI Officer Arturo LoAIza-Bonilla MD. Read the full story: lnkd.in/gBEZAJv2 @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee
Massive Bio@MassiveBio

@MassiveBio is collaborating with @OpenAI to expand global access to clinical trials. 🌍 Through our collaboration with @OpenAI, we're enabling real-time trial parameterization and free pre-screening services in high-population regions, ensuring that life-changing opportunities are available to patients everywhere, not just those near major research centers. This is what AI-driven precision oncology looks like in practice: breaking down the barriers that have long kept underserved patients from accessing the trials that could change their lives. Read the full announcement → massivebio.com/massive-bio-ex… #ClinicalTrials #AIinHealthcare #PrecisionOncology #HealthEquity #CancerResearch #OpenAI #DigitalHealth

English
2
2
6
1.6K
500 Global retweetledi
Sakana AI
Sakana AI@SakanaAILabs·
We’re excited to introduce KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI, accepted at #ICASSP2026! 🐢 Blog pub.sakana.ai/kame/ Paper arxiv.org/abs/2510.02327 Can a speech AI think deeply without pausing to process? In real conversation, we don’t wait until we’ve fully worked out what we want to say—we start talking, and our thoughts catch up as the sentence unfolds. Fast speech-to-speech models achieve this, but their reasoning tends to stay shallow. Cascaded pipelines that route through a knowledgeable LLM are smarter, but the added latency breaks the flow—they fall back to "think, then speak." In our new paper, we propose a way to break this trade-off. We call it KAME (Turtle in Japanese). A speech-to-speech model handles the fast response loop and starts replying immediately. In parallel, a backend LLM runs asynchronously, generating response candidates that are continuously injected as "oracle" signals in real time. This shifts the AI paradigm from "think, then speak" to "speak while thinking." The backend LLM is completely swappable. You can plug in GPT-4.1, Claude Opus, or Gemini 2.5 Flash depending on the task without changing the frontend. In our experiments, Claude tended to score higher on reasoning, while GPT did better on humanities questions. Try the model yourself here: huggingface.co/SakanaAI/kame
English
14
144
735
285.5K
500 Global
500 Global@500GlobalVC·
@DeepInfra's $107M Series B underscores a key shift in AI: inference is becoming the system constraint. As AI moves from training to production-scale deployment, high-throughput, reliable inference infrastructure is becoming foundational to real-world adoption—from agent-driven workflows to always-on enterprise applications. With this funding, DeepInfra is scaling its inference cloud to meet global demand, supporting 4T+ tokens processed weekly across 150+ open-source models and an expanding footprint of 8 U.S. data centers. We’re proud to co-lead this round alongside @gharik and an incredible group of investors supporting DeepInfra’s next phase of growth. Congratulations to @nikolaborisof, Yessenzhar Kanapin, Georgios Papoutsis and the whole DeepInfra team on this milestone. Full story: lnkd.in/g6vRFyEH And…they’re hiring! Take a look here: lnkd.in/d8pzKjtY @christine_tsai | @CourtneyPowell | @TonyW | @vishalharnal | @khailee | @dfect | @leoclee #AIInfrastructure #SeriesB
500 Global tweet media
English
1
3
6
971
500 Global
500 Global@500GlobalVC·
Featherless has closed a $20M Series A, co-led by AMD Ventures and @AirbusVentures, to scale its position as a neutral inference layer for open models — independent of hyperscalers, chipmakers, and proprietary ecosystems. As AI adoption accelerates, inference is becoming a key control point in the stack. Featherless is leaning into this shift by expanding its model library, building an open foundation for next-generation AI applications, lowering inference costs, and scaling enterprise deployments, making it significantly easier for developers to operationalize and scale with open models. Already the fastest-growing inference partner in the Hugging Face ecosystem, Featherless supports over 30,000 open models — a strong signal of where the market is heading. Congratulations @picocreator, Wesley George, and the @FeatherlessAI team! Full story here: featherless.ai/blog/series-a @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee @khailee @AriefJohan
PicoCreator - AI builder @ AIE 🇸🇬@picocreator

AI is concentrating, in a few countries, a few companies, a few chips. Open-source AI is the only real check on that. And it only works if the infrastructure to run it actually exists. Today, we're announcing our Series A, $20 M, to make this a reality.

English
1
4
14
4.2K
500 Global
500 Global@500GlobalVC·
"The future of AI is collective intelligence." - @hardmaru Congrats @SakanaAILabs on the beta launch of Fugu! 🐡🐡
Sakana AI@SakanaAILabs

We’re launching the beta for our new commercial AI product: Sakana Fugu 🐡, a multi-agent orchestration system! Blog: sakana.ai/fugu-beta Fugu hits SOTA on SWE-Pro, GPQA-D, and ALE-Bench, and has been our internal secret weapon. It dynamically coordinates frontier models, autonomously selecting the optimal agent combinations and roles for each task. Available as an OpenAI-compatible API, you can seamlessly integrate Fugu into your existing workflows with minimal changes. 🐟 Fugu Mini: High-speed orchestration optimized for latency 🐡 Fugu Ultra: Full model pool utilization for deep, complex reasoning Apply for the beta test here: forms.gle/BtKkhc2CfLKk1d…

English
1
10
37
16K
500 Global
500 Global@500GlobalVC·
Cross-border mobility is becoming increasingly digital, but navigating visas still remains fragmented, opaque, and slow. @IssaCompass is changing that with an AI-powered visa platform built for speed and predictability. The traction is clear: over 100,000 users, 10,000+ inbound users monthly, and $1M in annual revenue within just five months. Even more compelling, the platform has achieved a 99% approval rate for pre-qualified applications. Beyond visas, Issa Compass is embedding into the broader ecosystem supporting global mobility, from health insurance to payments to local AI-native communities. This kind of vertically integrated, AI-enabled infrastructure is what will power the next generation of global mobility. Full story here: natlawreview.com/press-releases… @vishalharnal @atxfrench @khailee @AriefJohan @feliciachiaxl
English
5
7
7
952
500 Global
500 Global@500GlobalVC·
AI workflows are only as powerful as the context they can access. Today, a significant portion of how teams think, plan, and execute still lives in formats that traditional AI systems struggle to fully interpret or act on. Whether it's diagrams, boards, or complex discussions on Claude, @LucidSoftware is integrating visual intelligence into AI workflows so teams can move from insight to execution in seconds and build on existing knowledge without losing momentum. This is a meaningful step toward more integrated, multimodal AI systems that reflect how work actually happens, and where the next wave of productivity gains will come from. Full story here: aijourn.com/lucid-software… @dmgrow @karlxsun @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee
English
0
0
7
1.4K

Keşfet

@aircall @vogentai @ScottChancellor @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee