Giannis Chatziveroglou

pretraining @cohere, CS and Math @MIT

[Sylvie Shi @sylvieshi00]

I’m hiring experienced ML engineers for my team to work on pre-training data and mid-training. We have ambitious plans, and you’ll play a key role in shaping the data mixture for large-scale pre-training runs. Please consider applying if you care deeply about data and its impact on model performance. Links are in thread!

[Cohere @cohere]

Introducing Cohere's first open-source coding model: North Mini Code

Small & efficient, designed for agentic performance and built for community input.

[Giannis Chatziveroglou @giannis2two]

Aye captain🫡

[Dwarak @DwaraknathG]

Extremely excited to present Command A+, our first sparse model!

I am very proud of the work we did to enable this model. We built our sparse training stack from the ground up over the past year with a lot of custom kernels, performance engineering to enable us to train large sparse models with a very small compute footprint.

We presented our work on a fully dropless FP8 kernel stack for sparse models at Nvidia GTC earlier this year. We are extremely ambitious and are marching towards a state-of-the-art model training stack, come join us if you are interested in pushing the frontier!!

Apply here - https://t.co/BhV5ffuarA
Our work here - https://t.co/o86oK5iD2E

[Cohere @cohere]

Introducing: Cohere Command A+

We’ve created our most powerful LLM yet, optimized it to run on as little hardware as possible, and released it open-source for all.

[Cohere @cohere]

Excited to share our work on production-ready W4A8 inference, now integrated in vLLM! By combining 4-bit weights (low memory) with 8-bit activations (high compute), we hit the sweet spot for both decoding and prefill — up to 58% faster TTFT and 45% faster TPOT vs W4A16 on Hopper.

[Ekagra Ranjan @EkagraRanjan]

Ever wondered how Speculative Decoding interacts with production MoE models?

Conventional wisdom: MoE + speculative decoding = too many experts to load, gains disappear.

Reality: MoE amplifies speculative decoding.

Checkout Cohere Blogpost: https://t.co/0iqZpYfXIN

[Reid Wiseman @astro_reid]

There are no words.

[Nick Frosst @nickfrosst]

@cohere transcribe

Sota open source transcription model running in the browser :)

Weights on @huggingface link below

[Cohere @cohere]

Introducing: Cohere Transcribe – a new state-of-the-art in open source speech recognition.

[Cohere @cohere]

We’re excited to announce our partnership with @RWSGroup, bringing Cohere’s frontier AI models to Language Weaver Pro - unlocking new enterprise‑grade translation capabilities.

Purpose-built for high-stakes environments, this integration empowers enterprises and governments to communicate seamlessly across languages and accelerate new opportunities for global collaboration and growth!

Learn more: https://t.co/Bt7S9EcOwc

[Cohere @cohere]

We’re excited to announce that we’ve signed a Memorandum of Understanding for advanced AI collaboration with Saab, a global leader in defense and security.

Together, we’ll explore groundbreaking AI partnerships for their aerospace platforms and deliver tailored solutions critical to Saab’s operations. 

Read more: https://t.co/yvHeXKqYfn

[Dwarak @DwaraknathG]

Hey all, I will be at GTC next week talking about all the work my team and I did on large-scale MoE training in JAX on GPUs! We decided early on to have a fully dropless training stack to avoid token dropping. (1/7)

[Aston Martin Aramco F1 Team @AstonMartinF1]

Welcoming @cohere to the team as our Official Generative AI Partner to help accelerate AI innovation.

Learn more: https://t.co/LUnkYAw9Em

[Joe Weisenthal @TheStalwart]

I'm honestly astonished that Citadel Securities published a rebuttal to the @Citrini7 piece. https://t.co/bBohvSMoib

[Giannis Kaklamanis @gianniskakl]

We are excited to share our new paper on Verifiable Aggregate Receipts (VAR) — a cryptographic primitive for auditable engagement counts that prevents inflation while preserving user privacy.

This is joint work with @wanggary142857, Jasleen Malvai, and @0xFanZhang at @YaleACL.

[Dwarak @DwaraknathG]

I will be at Nvidia GTC in March! @bharatvenki and I are gonna talk about all the systems work we do at Cohere. Come listen from me about all the custom kernel work we do for large scale LLM training on Hopper and Blackwell!! https://t.co/2SEbDtrIQk

[The Greek Analyst @GreekAnalyst]

I love discovering extremely talented young Greeks like @giannis2two in NYC building super cool stuff. So many hidden (people) gems out there. Who are some of the highest signal yet most underfollowed GR techies you know? Drop them below, help me bring the spotlight they deserve.

[Laura Ruis @LauraRuis]

Apply to do research with me on emergence of agency/planning in LLMs, out-of-context reasoning, understanding generalization from data, or propose your own direction!

Very excited to be mentoring this spring 💫

[Kyle Duffy @kyduffy]

@SahilBloom You should distinguish uncertainty of purpose from uncertainty of rewards

E.g. it should be certain that you'll show up and take a step every day; even if it's uncertain you'll be rewarded

Uncertainty of purpose should be extinguished, uncertainty of rewards should be expected

[Sasha Rush @srush_nlp]

Video and blog post of @SonglinYang4 explaining DeltaNet used in the newest Qwen-Next model

* https://t.co/6TvJxjd8dQ
* https://t.co/PkrAZmu5oc