CTO Letter from Bryan Healey, Co-Founder and CTO of Aiera

April 11, 2024: Before Q1 2024 earnings kicks off, Aiera felt it was important to share the recent strides we have made for better event audio quality and event connection reliability. Please see below for more detail in a letter from Aiera's Co-Founder and CTO Bryan Healey.

Bryan Healey: As we head into this upcoming earnings season, I want to let you know about some of the engineering work that we have done to ensure a smooth and seamless experience ahead:

1. Increased Redundancy on Event Connectivity

We are always looking for ways to improve the stability of our event connection. An infrequent but frustrating issue that we occasionally face is the “drop,” whereby a hiccup on the host side (especially on webcast capture), or an issue with the audio relay, can cause the connection to be briefly lost, which had required us to stop and reconnect. To combat this, we have implemented an improved deployment method, along with various audio capture improvements, that will increase the resilience of our system against this kind of issue, and ensure a more stable, uninterrupted connection.

2. Vastly Improved Text/Audio Alignment

Once a live event is over, we swap the raw real-time transcripts with human-verified transcription. This provides various enhancements, including correction to the structure of the text for readability and cohesion (such as paragraph and speaker reorganization), as well as correction to any audio or speech irregularities (such as removing repeated or jagged statements, correcting any audio gaps or quality related flaws, etc.). However, it can be surprisingly difficult to then align the raw audio with the newly human-corrected text. We leveraged a machine learning approach known as forced alignment, which uses probability scoring to find anchor points in an audio stream that match with individual transcript segments. However, this is imperfect; so, we have recently deployed a new novel approach to alignment that marries classical forced alignment with large language modeling to provide unparalleled alignment accuracy. We have begun rolling out this new approach in just the last few weeks, and will be in full-force for earnings!

3. Expanded Automatic Monitoring of Stream and Transcription Quality

Quality has always been our top priority, and we have teams that work tirelessly to ensure accurate connection detail, on-time and stable audio connection, and a quality, uninterrupted transcription. But keeping close watch on every event, especially during the busy earnings season (when hundreds of concurrent events is not unusual) can be a daunting human challenge. We have thus rolled out new intelligent monitoring using language modeling to augment our quality efforts, allowing our teams to work faster and more efficiently even during the busiest hours.

We understand just how critical earnings is for our customers, and we see every season as an opportunity to renew our commitment to quality and reliability, and as an opportunity to make our customers fall in love with Aiera all over again.

We are excited for the weeks ahead, and are ready to help you through this stressful time, together.

Bryan Healey

Co-Founder and CTO

Aiera

- As always , if you or your team is interested in trialing Aiera during this earnings season, click here to request a demo.