AssemblyAI

🎉 New blog post: 7 LLM use cases and applications in 2025 From automating customer support to analyzing audio data at scale, LLMs are unlocking new levels of efficiency across industries. In the article, we explore 7 powerful LLM use cases that are reshaping business operations: ✅ Audio data analysis for actionable insights ✅ Content creation that scales ✅ 24/7 personalized customer support ✅ Real-time language translation ✅ Sentiment analysis at scale ✅ Personalized education and training ✅ Advanced cybersecurity threat detection Read the full post here:

🎃 Spooky season is here, but you know what's truly scary? Your speech-to-text adding mystery guests who were never in the meeting. 👻 While others are still haunted by phantom speakers, we've made our industry-leading speaker detection even more precise: - 36% more accurate at identifying the right number of speakers - Fewer "phantom participants" showing up uninvited - Better at catching every real voice in the conversation These improvements are live now for all customers. 🦇

Instead of post-call analysis, agents can now get live support during conversations—leading to immediate results: 📉 20-25% reduction in average handle time 📈 30% improvement in first-call resolution ⏱️ 50% less after-call work How it works: real-time speech transcription feeds AI models that analyze sentiment, ensure compliance, and surface relevant knowledge instantly. Agents can focus on empathy and problem-solving while AI handles information retrieval. We just released a technical walkthrough explaining real-time agent assist and showing how to build your own system using AssemblyAI's streaming API: https://lnkd.in/era5Dzk8

We’re proud to partner with teams that share our commitment to quality, innovation, and measurable impact. 📈 Calabrio’s dedication to transforming every customer interaction into actionable insight makes them an exceptional partner to work alongside! 🤝

Our Applied AI engineers created a voice-controlled version of the game Snake, using our streaming speech-to-text API 🐍 Just say "up", "down", "left, "right" to move in the desired direction. The current high score is 240 -- think you can beat it? https://lnkd.in/e7ZX8h9p

Come be a part of AssemblyAI! We're currently hiring for two mission-critical roles: Lead AI Platform Engineer: This is a technical lead role for an engineer who wants to drive the technical direction of our AI infrastructure. The ideal person will own the design and architecture of our data and AI training platforms, enhancing them to dramatically improve research velocity and outcomes. They should have deep expertise in distributed systems and be passionate about building scalable platforms with tools like GCP, PySpark, and Airflow. Senior Research Engineer: This is a highly skilled engineer who acts as a bridge between research and production engineering. This person will be embedded within our Research team, helping to diagnose and resolve performance bottlenecks across large-scale distributed training and inference systems. We need someone with a deep understanding of modern deep learning systems who can translate research prototypes into highly optimized, production-ready models. Experience with PyTorch internals and low-level programming (C++ or Rust) is a major plus. Apply here: https://lnkd.in/dq5V4R7t

Last week, our team got to spend the day at VapiCon! We're feeling energized after connecting with the voice AI community in San Francisco - it's incredible to see how much this space has grown. A huge thank you to everyone who attended our CEO Dylan Fox's talk. Here are some key takeaways he shared: 📈 The Voice AI Opportunity is Massive: The market is expected to grow 20x over the next decade, with over $3B in VC funding flowing into voice AI in just the last 20 months. We've gone from being one of the only voice companies in YC back in 2017 to voice AI representing 20% of YC companies today. 🎯 Accuracy is the Main Bottleneck: As voice AI applications scale into critical real-world scenarios, accuracy becomes paramount. We're focused on moving beyond simple speech recognition to true speech understanding—detecting emotion, handling noisy environments, managing background speakers, and delivering context-aware responses. 🏗️ The Infrastructure Ecosystem Has Matured: What you can build today as a small team is exponentially more powerful than even 18 months ago. The combination of robust APIs, LLMs, and supporting infrastructure means developers can now create production-ready voice applications faster than ever. The future of voice AI is incredibly bright, and we're thrilled to be building alongside this amazing community. If we missed you at VapiCon, reach out—we'd love to hear what you're working on!

From 41 escalated cases to 80% CSAT improvement in 90 days. Calabrio needed a change. Legacy speech tech was holding back their conversation intelligence platform. Outdated models meant poor accuracy, frustrated customers, and limited global expansion. The results after modernizing their transcription infrastructure with Assembly: 😃 80% boost in customer satisfaction 🚀 22% revenue increase ⚡ 96% faster response times 📉 18% reduction in error rates In workforce management and conversation intelligence, transcription accuracy directly impacts every downstream feature. When your Speech AI foundation is strong, everything else accelerates—from product innovation to international growth. Link to the whole story below ⬇️

We’re at VapiCon today! Come say hi and order a coffee through our voice agent ☕️

Congratulations LiveKit team - we are proud to be a partner! 👏