Accelerating LLM Inference with Speculative Decoding - LinkedIn Engineering Blog
LinkedIn's Hiring Assistant team shares how they used speculative decoding to achieve 4Ă— higher throughput and 66% latency reduction for their AI agent. This deep dive explores n-gram speculation, implementation strategies with vLLM, and practical lessons for serving LLMs at scale in production.
Nov 6, 2025
The Real AI Bubble Is in Data Centers No One Can Power Up - HackerNoon Article
HackerNoon feature by Dhyey Mavani exploring how the real speculative excess in artificial intelligence lies not in the models, but in the physical infrastructure powering them. The article examines the mounting risks behind the rapid expansion of GPU data centers and energy-hungry AI compute projects that face power, supply chain, and efficiency constraints.
Oct 27, 2025
Agents In The Loop Hackathon - 2025
A weekend hackathon in San Francisco focusing on building agentic workflows that work with us and not (only) for us, featuring $5,000+ in prizes sponsored by [Anthropic](https://www.anthropic.com/), [Google](https://www.google.com), [Hudson River Trading](https://www.hudsonrivertrading.com/), [Entrepreneurs First](https://www.joinef.com/), [Mastra](https://mastra.ai/), etc. and expert judges from leading tech companies.
Aug 2, 2025
🚀 Key Learnings from Y Combinator AI Startup School 2025
Insights from OpenAI’s Sam Altman, Microsoft’s Satya Nadella, Andrej Karpathy, Elon Musk, and others on the evolving AI landscape, startup defensibility, agent-based systems, and next-gen product development.
Jun 17, 2025
Y Combinator AI Start-up School 2025
Selected and sponsored to attend Y Combinator’s inaugural AI Startup School, joining 2,500 global builders and leaders in San Francisco.
Jun 16, 2025
Panel 2 - AI Applications - LogFlowAI - Transforming DevOps with Automation, Real-Time Analytics & AI
Panelist for AI Applications Representing LogFlowAI initiative (proactive log-based analysis instead of reactive approaches)
Apr 19, 2025
Hack the Herd 2025 - AI in Sustainability Hackathon
A 24-hour hackathon focusing on AI-driven solutions for campus sustainability challenges, featuring $1000+ in prizes and industry expert judges.
Mar 29, 2025
TreeHacks 2025 - LogFlowAI's Journey from Concept to Reality
Selected as one of 1000 hackers from 12,000 applicants, transformed HackMIT Y Combinator winning concept LogFlowAI into a functional MVP in 16 hours.
Feb 14, 2025
DeepSeek and the Next Wave of AI Innovation - A Critical Discussion
The event was headlined by chief guest Dhyey Mavani, whose extensive industry experience from Amazon and LinkedIn—working on cutting-edge AI/ML models and GPU infrastructure—set the stage for an insightful exploration of today's AI landscape. Dhyey emphasized AI's role as an augmentation tool, rather than a replacement for traditional, seminal technologies, detailing how it can enhance recommendation engines, operational management, and workflow scalability.
Feb 11, 2025
Matching Engine
The Matching Engine is an end-to-end ML-based algorithm surrounded by Django-based, fully integrated infrastructure. It efficiently matches teams for i2i startup and pitch events based on participant responses compiled from a Google form.
Sep 1, 2024