DevAd at @SimpliSmartHQ 🥑 Inference magician ✨ Mozillian @mozilla 🦊 Ex. @yellowdotai, @zyte, @intel #DevRel #Edu
C-137
Joined September 2016
- Tweets2.2K
- Following2.6K
- Followers912
- Likes16.7K
Mom asked me to count the modak
I pulled up ChatGPT and got it done in 10 seconds.
End result: 5 min bhaashan from mummy about how our generation is spoilt 🥲
Ab isme kya galat kiya maine 😭
Has anyone tried model hot-swapping or sleep-wake with @vllm_project in production?
Particularly curious about wake-up time, CPU RAM pressure, TP, request draining and edge cases
What broke or surprised you?
My next conference talk just got a whole lot cooler 🤩
All thanks to @harshil1712 @typesafeai
What if your slides are smart enough that they navigate automatically? 👀
Thank you @typesafeai for providing access to Jev! I had this idea for a few weeks, and Jev felt like the right model for this task.
(Sorry for the bad audio)
IMAX me Odyssey dekhne aaya tha, par yahan toh pure Nexus Shantiniketan mall ki bijli hi chali gai 🫠
Kya itna manhoos hu kya? 🥹
Want to know what exactly happens inside an inference request?
@rudrakshkarpe @HowDevelop bhai aaye hai, btake jayenge!
Vibe coding is dead. @mesirii at @agentsnexus is calling it what it actually is: agentic engineering 😎
What the duck!!!!
BIG ANNOUNCEMENT FROM HUGGING FACE TODAY:
We're unveiling Microduck 🐥🤖
It's a tiny $399 open-source robot you can teach new tricks with reinforcement learning. It can walk, pick things up, get back up when it falls, and even roller-skate.
Welcome to the era of open-source affordable robots to democratize physical AI and world models!
🤗🤗🤗
Pratik Parmar retweeted
Hello Vadodara folks 👋
I'm hosting a small AI co-working session this Saturday.
If you're building with LLMs, AI agents, MCPs, AI tooling or anything interesting in the AI space, come join us.
No fixed agenda, just a bunch of builders getting together to co-work, exchange ideas, network or demo whatever we're working on.
Luma link ⤵️
Pratik Parmar retweeted
Hello Vadodara folks, I am thinking of organizing a small co-working session for people building cool things around AI agents, AI tooling, or anything around LLMs.
The core idea is to meet, showcase what you’re building, and exchange ideas. If you’re interested, hmu. If we get enough people, we will set smthg up!
I was today years old when I noticed @posthog has a hedgehog companion just to prove that analytics doesn't have to be a slog.
Max is doing more for my motivation than any dashboard metric. 🦔
Pratik Parmar retweeted
🚀 We're excited to be a Day-0 launch partner for NVIDIA Nemotron 3 Ultra.
Deploy NVIDIA's latest open model for agentic AI on Simplismart
Our optimizations we deliver higher throughput than TensorRT + MTP + NVFP4.
Read more: lnkd.in/g7BrrNhy
#NVIDIA #Nemotron
Going live tomorrow to build a production-ready AI voice interviewer agent 🎙️
Sub-400ms latency, open-source models on @SimplismartHQ , voice agent orchestration with @LiveKit
5:30 PM PT 👉 luma.com/jg93tpel
I’ve been playing around with Qwen3-TTS for a real-time voice agent I'm building and it just blows my mind how insanely low the TTFB is 📉
When you're building conversational pipelines, latency is everything.
Here is what makes this TTS model so special 🧵 (1/n)
Instead of raw audio, Qwen uses a discrete multi-codebook tokenizer. It compresses sound into layers of discrete codes representing semantics and acoustics.
To clone a voice, it just needs a 3s reference audio to create an acoustic prompt blueprint. (2/n)