PhoenixAI (formerly CelerData)

Some months you ship a feature. This month we shipped a whole new identity. 👉 The May issue is out — the rebrand story and the strategic bet behind it, a first look at Agent Fawkes (the new AI copilot inside PhoenixAI Cloud) plus what else landed this month, the latest stories of how customers are running agentic analytics in production, and our June 11 launch webinar on where we're going next. 🐦🔥 May the force be with PhoenixAI. 🐦🔥 The full recap 👇 #PhoenixAI #ai #data #analytics #database #dataengineering

You’ve heard the name. Now come see what PhoenixAI is all about. 🐦🔥 On June 11, we’re hosting a live session - 🔗 https://lnkd.in/e--KVRN7 with Sida Shen to share what PhoenixAI is, why we’re building it, and how it’s designed for the next wave of AI applications and agents. Sida will walk through what’s available today in PhoenixAI Cloud, where we’re headed next, and how PhoenixAI supports the full agent loop — from decision to execution to trace. We’ll also leave time for live Q&A with the team! #PhoenixAI #ai #data #analytics #database #dataengineering

LLM handles the conversation. PhoenixAI handles the truth. 👉 That's how Conductor describes the split-reasoning architecture behind their agentic analytics — and why their AI agent returns decimal-level accuracy on customer-facing data, every run. Most MCP implementations hand raw data to the LLM and let it do the math. The math goes wrong. Even an 80% accuracy rate destroys trust when customers (Zoom, Citi, Louis Vuitton, Pfizer) are making seven-figure marketing decisions on the output. Conductor's fix: the LLM interprets intent and picks tools. PhoenixAI executes the joins, aggregations, and arithmetic on terabytes of data — sub-second. The LLM never touches arithmetic. The results: ▸ Sub-second queries on 180–240M-row datasets (389ms–2.8s) ▸ Idea to production in <2 months on a new flagship product ▸ 100% decimal accuracy, deterministic across runs Read the full architecture 👉 https://lnkd.in/gDza2f3u #ai #data #analytics #database #dataengineering #realtimeanalytics #aiagent

⚡ Real-time ad targeting. 3× faster BI dashboards. An internal AI analytics agent that hits 90%+ accuracy on real business questions. 👉 CTW — the Tokyo-based team behind G123, one of the world's largest HTML5 browser gaming platforms, serving anime & manga IP to players in 200+ countries — rebuilt the serving layer behind audience segmentation, BI, and AI analytics on PhoenixAI. The results: ✅ Audience segment refresh: hours → seconds (<10s data freshness) ✅ Tableau on Delta Lake: 3× faster dashboards under concurrency ✅ Doraemon, their multi-agent analytics assistant: 90%+ answer accuracy 💬 CTW's takeaway: The bottleneck wasn't whether TiDB or the lakehouse could keep up — it was whether either was the right shape for the workload. Read the full story 👉 https://lnkd.in/gs3ypwdV #ai #data #analytics #database #dataengineering #realtimeanalytics #aiagent

Still taking in what a big week this has been. 🐦🔥 PhoenixAI (formerly CelerData) is more than a new look. It marks a clearer mission, a sharper category focus, and a bigger chapter for the company. AI agents are changing the way enterprises use data. They need live access, real-time performance, massive concurrency, and deployment options that meet production requirements. At that scale, the database can either become the bottleneck or unlock what’s possible — and PhoenixAI is built to be the breakthrough. And with Rick Underwood joining as President, this next phase starts with new leadership and real momentum Onward. 🔥 Read the story → https://lnkd.in/gdfvruw4 See the rebrand → https://lnkd.in/geiQkNhe #PhoenixAI #ai #data #analytics #database #dataengineering #realtimeanalytics #aiagent

🥂 Meet us at Ula Restaurant in San Francisco on Monday, June 2 (6–8pm). We're hosting a small, invite-only happy hour for a curated group of data + AI leaders. 🎟️ RSVP: https://lnkd.in/gud35yEn No crowds. No chaos. No badge scans. Just great beverages and conversations that are actually worth your time. If you're in town and want in, RSVP below. Space is limited and approvals are by hand 👇 🎟️ https://lnkd.in/gud35yEn P.S. If you're in SF for Snowflake Summit week, even better — come unwind with us! #PhoenixAI #ai #data #analytics #database #dataengineering

Rick Underwood is here. Read his take on the shift behind PhoenixAI (formerly CelerData) 👇 🐦🔥 #PhoenixAI #ai #data #analytics #database #dataengineering

⚡ PhoenixAI serves AI agents the way they actually query: ad-hoc, sub-second, at thousands of concurrent QPS on live enterprise data. The Real-Time Analytical Database for AI agents. In your cloud, or self-hosted. 👉 See the product: https://lnkd.in/gJk4UBPK #PhoenixAI #ai #data #analytics #database #dataengineering

CelerData is now PhoenixAI.🔥 The Agentic AI Database. Built so AI agents can query live enterprise data at sub-second latency and massive concurrency — in your cloud, or self-hosted. Most analytical databases were architected for a world that no longer exists. Humans ran dashboards on flat tables. Complexity was someone else's problem. That world is over. When thousands of agents need to query, reason, and act on petabytes of live data at once, the database is either the bottleneck or the breakthrough. PhoenixAI is the breakthrough. Built on three things enterprises need to put AI into production: ⚡ Speed. Sub-second latency at agentic concurrency, so agents act on what's true now, not yesterday's snapshot. 🌊 Simplicity. One engine for real-time and historical data. No brittle pipelines. Lower total cost of ownership. 🔒 Governance. Data sovereignty, RBAC, SOC 2, encryption. Deployed in your own cloud (BYOC) or self-hosted with PhoenixAI Anywhere. Already in production at Intuit. Expedia. Pinterest. Also new this week: Rick Underwood (ex-Snowflake, ex-CEO of Clumio) joins as President. Welcome to PhoenixAI. 🔥 👉 Check out our brand new website: https://lnkd.in/d-h5xhVE  #PhoenixAI #ai #data #analytics #database #dataengineering

📚 Serving 100M+ customers means scale isn’t optional. Intuit’s IPS now delivers 100k+ events/sec, TP99 <500ms, and end-to-end freshness <4s—so Intuit Assist and support models can act on the last 30 clicks while they still matter. By moving OLAP to StarRocks, the team: • Hit a strict <4s freshness SLA to protect model accuracy • Sustained bursts >100k TPS while keeping TP99 under 500ms • Eliminated denormalization workarounds with fast multi-table joins • Achieved ~2s end-to-end ML latency (1s freshness + <500ms query) If you care about real-time analytics without operational sprawl, this story is a great read 👉https://lnkd.in/gfgQCazY #analytics #dataengineering #database #datalake #realtimeanalytics