About
- Visionary and experienced leader in the Data & AI/ML field
- Passionate about…
- Passionate about…
Activity
9K followers
-
Chen Peng reposted thisChen Peng reposted thisToday we’re launching AutoEval: a new evaluation methodology that ranks models using reward models based on millions of real Arena user preferences. Highlights: - High-quality evaluation signals calibrated on real preference data - Strong alignment with live human evaluations - Evaluations that are several orders of magnitude faster (hours instead of days) - Support for Text, Vision, Image, and Code Arena AutoEval enables us to evaluate newly launched models much faster and share results with the community sooner. We’ll now show AutoEval estimated scores for new models directly on the leaderboard. More details in the article.
-
Chen Peng shared thisIncredibly excited and proud of this major milestone---$100M annualized revenue--that the small but mighty Arena team achieved in 8 months! We are just getting started. Join us: arena.ai/jobsChen Peng shared thisArena reached a $100M annual revenue run rate just 8 months after launching our evaluation product. We started as a research project at UC Berkeley with a simple mission: measure AI progress through real-world use. As AI shifts from chatbots to agents taking on longer, higher-stakes work, the problem matters more than ever. Today, Arena measures real-world AI utility with a community of tens of millions. With Agent Arena, we’re evaluating long-running agents on complex, real-world tasks - how they use tools, adapt to feedback, and recover from errors, and accomplish goals set by humans. We are excited to keep deepening our work in agentic evaluations. Here’s Anastasios Angelopoulos on what this milestone means and where we go from here: If our mission resonates with you, check out our job openings and reach out: arena.ai/jobs If you’d like to learn more about the milestone, head over to our blog: https://lnkd.in/g83eynhg
-
Chen Peng reposted thisChen Peng reposted thisIntroducing Agent Arena: real-world agentic evals at scale. How do you evaluate agents doing actual work? We measure millions of live sessions where real users accomplish real tasks. On Arena, models now get web search, filesystem, and terminal tools to complete complex workflows: writing code, creating slide deck, researching the web, building apps, and analyzing documents. Every session produces rich signals. Users iterate with the agent turn-by-turn: approving, editing, correcting, praise or expressing frustration. The environment gives feedback too: shell errors, tool failures, recovery attempts, and more. Our leaderboard measures each model's agentic performance using causal inference across five signals: task success, steerability, error recovery, user praise vs. complaint, and tool hallucination. This leaderboard snapshot is built from 300K+ tasks, 2M+ tool calls, and 40M lines of code by agents. Top labs in Agent Arena: - #1 OpenAI: GPT-5.5 (High) - #2 Anthropic: Claude-Opus-4.7 (Thinking) - #3 Z.ai: GLM-5.1 - #4 Google DeepMind: Gemini-3.1-Pro - #5 Kimi (Moonshot AI): Kimi-K2.6 More analysis in the thread, with the full technical blog below. What are people actually using agents for? We analyzed the task distribution in Agent Arena across a 7-day window: 160K real user tasks spanning coding, debugging, research, document creation, frontend development, file analysis, and long multi-step workflows. The largest categories were: - Code writing (17.5%) - Research and lookup (10.8%) - Planning and brainstorming (10.6%) - Multimodal image/video work (10.2%) - Document creation (9.1%) - Code debugging (8.9%) Agent usage is broad: it’s not just coding, but research, planning, content creation, file work, and complex workflows that combine multiple tools over many turns. The aggregate ranking combines multiple signals: task success, user praise vs. complaints, steerability, bash recovery, and tool hallucination. Top models win in different ways: some complete tasks more reliably, some recover better from errors, and some are easier for users to steer. Higher-cost models generally deliver stronger agentic performance, but not always. Agent Arena helps measure the trade-off: which models are strongest, which are most efficient, and track how the frontier is moving. In a 7-day window, Agent Arena logged 2.06M tool calls across 160K+ real user tasks. The most-used tools were: - bash: 936K calls - write_file: 550K calls - web_search: 276K calls - read_file: 118K calls - fetch_page: 86K calls This gives us a new lens on agent behavior: not just what models answer, but how they search, code, edit, recover, and interact with the environment. This is the core motivation for Agent Arena: evaluate agents on real, messy, long-horizon work. Check out our technical blog for the Agent Arena methodology: https://lnkd.in/gjJ6yQNc The full Agent Arena Leaderboard is here: arena.ai/leaderboard/agent
-
Chen Peng reposted thisChen Peng reposted thisYou can’t reliably evaluate AI on subjective qualities - like creative writing, tone, or whether a response is genuinely helpful - using a multiple-choice test. That's the core problem Arena solves. Instead of static benchmarks, we use live human preference data, millions of pairwise votes from real users, to rank models based on how they perform in real world use. In this clip, our co-founder and CEO Anastasios Angelopoulos walks through why human evaluation matters and how the Arena leaderboard is calculated using Bradley-Terry modeling and logistic regression on vote data. The full talk covers the complete methodology, from the math to prompt-specific adaptive leaderboards. Watch it YouTube: https://lnkd.in/gjGamPsM
-
Chen Peng reposted thisChen Peng reposted thisToday, Arena is announcing our Academic Partnerships Program, a new initiative to support independent academic research in AI evaluation, rankings, and measurement. As AI systems advance and adoption accelerates, the methods we use to evaluate and compare models increasingly shape both scientific progress and real-world outcomes. Many of the most important contributions in this area come from the academic research community, and we’re proud to help support that work directly. Selected projects may receive up to $50,000 in research funding. We welcome proposals across evaluation methodology, leaderboard design, measurement and statistical validity, preference data and human evaluation, and safety/alignment evaluation. ◾ The Q1 submission deadline is March 31, 2026. We aim to respond within 8 weeks following the deadline. ◾ Read more our blog: https://lnkd.in/gyxctgna ◾ Apply here: https://lnkd.in/gdPrQvuU
-
Chen Peng shared thisExciting news to kick off the new year: I’ve recently joined LMArena as Head of Data & Applied ML, where I’ll be helping build the platform that measures — and accelerates — the frontier of AI for real-world use. Over the past year, LMArena has rapidly become one of the most impactful evaluation platforms in the world. We now serve tens of millions of monthly visits and have grown to well over $30M ARR, as millions of consumers access frontier AI models through us for free and model developers rely on us to understand model quality and deploy AI safely and effectively. Today, we’re also announcing another major milestone: $150M+ raised at a $1.7B+ valuation — a huge validation of the momentum the team has built and the massive opportunity ahead. Real-world, trustworthy evaluation is becoming foundational infrastructure for the entire AI ecosystem, and I couldn’t be more excited to join Anastasios Angelopoulos, Wei-Lin Chiang, Ion Stoica, and the rest of this extraordinary team to push the frontier forward. We’re hiring across the board — full-stack engineers, data/ML scientists & engineers, product analysts, researchers, and more. If you want to work at the center of the AI ecosystem — where models meet the real world — DM me or drop a comment. Let’s build something generational together! #ai #data #ml #evaluation #hiringChen Peng shared thisArena has raised $150M+ at a valuation of $1.7B+. 💪🏼 In the past 7 months, LMArena has: (A) Grown our userbase 25x. 35M+ unique users. (B) Grown our revenue from 0 to >>$30M+ ARR in 4 months. Our products help labs and enterprises measure the real utility of AI and understand their strengths and weaknesses for real users. (C) Grown our team to 40+ world-class experts in machine learning, product engineering, design, marketing, BD, and more. We are looking for world-class ML scientists, engineers, marketers, and more. Evaluation is one of the hardest and most important problems in AI, and we need brilliant methodologists and builders to help. If our mission resonates with you, apply to join us. My DMs are open and I will personally look at every message I receive. Come work side-by-side with me, my cofounders Ion Stoica Wei-Lin Chiang, and our amazing team solving important new challenges in research and product every day. Shape the future with us; all types are welcome, from academics to founders. We are building one of the fastest growing businesses in the world. Our mission is to measure and advance the capabilities of AI for real users. New product experiences are coming on LMArena for our community. New analyses and evaluations are coming for labs and enterprises to improve their AI systems based on real feedback. We are so grateful to all our users and customers and excited to serve you! The fundraise was led by Felicis and UC investments, with participation from Andreessen Horowitz, The House Fund, LDV Partners, Kleiner Perkins, Lightspeed, and Laude Ventures. Our whole cap table is digging in after seeing our growth. We are also excited to invite Peter Deng, GP at Felicis and former VP of Consumer Product at OpenAI, and Jagdeep Baccher, CIO of UC investments, to our board. Along with the amazing Anjney Midha who incubated us from the early days and cofounders Wei-Lin Chiang and Ion Stoica, we are building the strongest team in the world to solve AI evaluations for reliable deployment. Onwards!
-
Chen Peng shared thisA bittersweet announcement: After 4.5+ incredible years, I’ll be bidding farewell to Faire to take a planned career break before exploring what’s next! I’m deeply grateful to Max Rhodes, Daniele Perito, and Marcelo Cortes for giving me the opportunity to join this journey four and a half years ago. Building a technology-powered marketplace that empowers millions of brands and retailers worldwide to grow their businesses has been an inspiring mission, and I feel honored to have been part of it. A huge thank-you to all the amazing leaders I’ve had the privilege to work with and learn from over the years. Your guidance, support, feedback, and encouragement have been invaluable. It’s hard to imagine a stronger group of co-conspirators to tackle challenges alongside :) And last but certainly not least, to my beloved team: We started as just a handful of people and grew to 70+ strong, building and scaling some of Faire’s most critical data and AI/ML systems—from data infrastructure and the AI platform to search and personalization, pricing and incentives, underwriting, shipping, listing quality, marketplace quality, and more. Our work has had a profound impact on our customers’ lives and has been instrumental in driving Faire’s growth to where it is today. I couldn’t be prouder of what we’ve accomplished together, and I’ll be cheering for you from the sidelines, eagerly waiting to see your next achievements! Onward to the next chapter!
-
Chen Peng reposted thisChen Peng reposted thisWe're Hiring! Faire is looking for a talented AI Platform Engineer to lead the design and execution of our AI Platform, which powers the future of our wholesale marketplace. You'll architect and build scalable, reliable systems that enable seamless AI/ML deployment, driving core metrics for Faire's growth. Great opportunity to work alongside Namit K. in a high-impact role. 🚀 Apply using the link below or message me directly to get things going! https://lnkd.in/g26zBYQv
-
Chen Peng reposted thisChen Peng reposted thisRunning Faire would not be possible without data. Data powers hundreds of models and experiments that directly improve onboarding, search, and checkout workflows, as well as thousands of dashboards and data pipelines used for analytics. When it came time to move to a new data replication platform, we knew we needed something that could handle the scale of our operations. Senior Data Engineer Aditya Joshi shares our data replication journey, including what we learned from migrating replication infrastructure and how we’ve benefited from moving our sync infrastructure to Fivetran: https://lnkd.in/gJrD4Pd4Reliable data replication: What we learned from migrating to FivetranReliable data replication: What we learned from migrating to Fivetran
Chen Peng commented on a post
3d
Congratulations! 🎉
-
Chen Peng liked thisA new start for Gemini.Chen Peng liked thisToday we’re introducing Gemini 4 Argon, our next era of frontier intelligence. It delivers frontier performance in complex workflows across real-world software engineering, knowledge work, and cybersecurity defense with an industry-leading 1M token output limit. Gemini 4 Argon is rolling out to an initial cohort of cyber defenders through our Fairwind Program so they can leverage its full frontier-level cybersecurity defense capabilities. We'll continue to gather feedback from early testers as we iterate on guardrails before making Argon available to developers, enterprises, and consumers as soon as possible. Learn more → goo.gle/4AZLPRt
-
Chen Peng liked thisChen Peng liked thisThe world of AI assistants is absolutely exploding. It was a blast joining Anish Acharya on The a16z Show to talk about where this space sits today and where we believe it is heading. We broke down a number of different topics: - Poke, Instinct, Muse: the agent boom - What Assistant Benchmark actually tests - Cost savers beat time savers - Muse charm as Meta's data play - Silent agents in group chats - Proactivity is the real moat - Assistant vs agent, defined - Amazon blocks Muse, Shopify opens the door - The $20/day agent economics If anyone is ever interested in seeing how these different assistants compare to one another, check out Assistant Benchmark (link in comments) Comment any questions below and I'm happy to share my learnings with whoever is interested.
-
Chen Peng liked thisHonored to be part of this mission.Chen Peng liked thisFounded in 2026, FireWERX serves one mission: accelerate innovation and empower the Firefighter to end the wildfire crisis. After some of the most devastating fires in California's history, a Wildfire Innovation Ecosystem of mission-driven funders, researchers, academics, technologists, and policy experts came together around one question: How do we accelerate innovation to enhance fire agency capabilities through a new fire innovation hub? Our work starts with Fire Innovation Units (FIUs), which are dedicated innovation programs within local, state, federal, and international fire agencies focused on identifying innovation needs and rapidly evaluating and adopting the most promising solutions. We connect FIUs to the broader Wildfire Innovation Ecosystem to move promising solutions from pilot to scale. This is a true moment of alignment and the world is paying attention. FireWERX is dedicated to coordinated action and ensuring we don’t lose the opportunity in front of us. No agency can do this alone, and now no agency has to. Thank you to those who helped tell our story and define the road ahead: Scott Gregory, MBA of California Department of Forestry and Fire Protection (CAL FIRE), Mike Sheehan of Orange County Fire Authority, Matt W. of Megafire Action, Genevieve B. of Gordon and Betty Moore Foundation, Patrick Roberts of RAND, Gemma Guilera, PhD of Stanford Doerr School of Sustainability, Aida Baldini and Tatiana Molina of Chile-California Council. Please watch this short video, and share it with someone who should be part of this work. #WildfireResilience #FireWERX #FireInnovationUnit #FIU
-
Chen Peng liked thisChen Peng liked thisTwo months ago I said the easiest way to use open models in Codex was `brew install baseten-switch`. Starting today you don't need baseten-switch: Baseten is the first inference provider in the new OpenAI B2B Marketplace. Enterprise OpenAI customers can run open models on Baseten natively via Codex and the Responses API. https://lnkd.in/gBA7_HPk
-
Chen Peng liked thisChen Peng liked thisThank you Newsweek for recognizing Faire as one of America’s most admired workplaces, with a five-star rating! https://lnkd.in/ge4vudx6 We work hard on behalf of our retailers and brands, and this mission is something that really sets Faire apart. What’s more, we work hard on behalf of each other to ensure Faire is a great place to work, where every employee can make an impact and grow in their careers. If you’re looking for an opportunity to make a real difference in communities around the world, check out the open positions on our page!
-
Chen Peng liked thisChen Peng liked thisWorld Labs is joining AMD. This is a huge moment for World Labs, our team, and for me. I wanted to take a moment to share what this means and why I’m so excited for this next chapter – read more in my Substack linked below. The research and technical breakthroughs we have achieved since our founding in 2024 have given us a clear vision for AI’s potential to solve problems in the spatial and physical world. To accelerate into this future requires scaling our efforts, scaling our reach, and getting closer to the hardware. We began a deep technical partnership with AMD last year, starting with model training and inference optimization on AMD GPUs. As our teams worked together, we realized it would be a natural fit to bring together our AI ecosystem of software and hardware, foundation models, and applications. I will join AMD as an Executive Vice President and Chief Scientist, working directly with CEO Dr. Lisa Su, Justin Johnson and Ben Mildenhall to continue leading the World Labs team as it joins AMD to form a world leading frontier research organization. Together, we are committed to building out an end-to-end open AI ecosystem spanning hardware, software, platforms, and widely accessible open models. https://lnkd.in/g8gQ54aj
-
Chen Peng liked thisChen Peng liked thisWe're hiring at Arena for a few exceptional product leaders who are hungry to push the frontier of AI intelligence across a few areas. If you know of someone, or are interested, please reach out. https://lnkd.in/gkK26Krs
Experience & Education
Volunteer Experience
-
Co-founder / Former President
China America Innovation Network
Science and Technology
Publications
-
Capacity Planning in the Semiconductor Industry: Dual-Mode Procurement with Options
Manufacturing & Service Operations Management
Patents
-
Multi-Layer Optimization for a Multi-Sided Network Service
Issued US 20200074523A1
See patentMulti-objective ranking system for Uber Eats consumer app
Courses
-
Control Theory
-
-
Convex Optimization
-
-
Data Mining
-
-
Data Structures and Algorithm Analysis
-
-
Database
-
-
Dynamic Programming
-
-
Econometrics
-
-
Fundamentals of Software Development
-
-
Game Theory
-
-
Introduction to Artificial Intelligence
-
-
Investment Science
-
-
Linear/Nonlinear Programming
-
-
Mathematical Modeling
-
-
Microeconomics
-
-
Monte Carlo Simulation
-
-
Operations & Supply Chain Management
-
-
Programming in C/C++
-
-
Signals and Systems
-
-
Stochastic Process
-
Languages
-
Chinese
Native or bilingual proficiency
-
English
Full professional proficiency
Recommendations received
25 people have recommended Chen
Join now to viewOther similar profiles
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content