Skip to main content
Publish date:

Industry Insider: Musiio CEO/Co-Founder Hazel Savage

We talk to the co-founder of Musiio, an AI music company that searches, tags and categorizes music to make it easier to find and use.
Hazel Savage

Hazel Savage

Categorizing music can be a difficult business. Music can often defy boundaries and traditional terms, while musicians never like to be boxed into a few simple terms that will define them. However, it can be very useful for practical purposes when finding them in a sea of music that is released daily, weekly and monthly. It is said that 40,000 new songs are uploaded to Spotify each day and trying to get through everything that is released is simply impossible. There isn’t enough time in a day to listen to all of the music that is made and distributed across various music platforms. That is a problem one can't solve, but one can help mitigate. Musiio is a company that wants to help make it easier for people who are searching for types of music like music supervisors.

It is a company that uses AI (more on that later) to scan music for its genre, BPM, key, mood, tempo, emotion, instruments and energy. Using its search feature, you see those characterizes in a song that you input to their algorithm or you can search by a certain mood, key or BPM. This is useful if you are searching for something upbeat, but don’t know exactly what song you want or want to find a lesser known track that may not have been on your radar already.

To learn more about the Singapore-based company, we chatted with its CEO and co-founder Hazel Savage for a new Industry Insider feature. She got her start at HMV in London and then has snaked her way through the business working at Pandora and Universal before co-founding Musiio in 2018. We talk about the application of the company, roadblocks she faced in the business and some of the myths of AI in music.

Musiio Team

Musiio Team

How did you get in the business?

I started working at HMV in Darlington whilst at University, and then went full time with HMV in London. Once in London I was applying for jobs in Music and I landed Shazam. Next thing you know 14 years later I am still in the music/tech industry!

Did you find it easy to transition from one part of the music business to another?

I think if you are naturally curious then you can excel in any line of work, applying what you know where you can, and learning the rest. Everything I know from marketing to APIs I haven't learnt “on the job.”

What was one of the biggest roadblocks you had to overcome as someone trying to make their way in the music business?

The biggest challenge was that “first job” when you have no experience. I couldn't afford to do an unpaid internship so I had to hustle (and wait) for an opportunity at entry level that also paid. I applied for over 150 roles over a year while I was at HMV before I got the interview at Shazam. It was also the only interview I got. So when I got the interview I made every second count!

What traits and qualifications do you look for in potential new employees?

I have hired a lot of people over the years, and I look for a few recurring traits. Attention to detail: what do they do without being prompted. Hustle and Grit: Does this person go the extra mile and seem resilient to challenges? Joy/Energy: Are they a good person to be around? Are they trustworthy? Most skills can be taught but the above are how I select people I hire.

What was the biggest challenge of leaving established companies and starting your own? How have you navigated the investor fundraising cycle?

One of the biggest challenges is probably evidenced by the fact I waited until I was 37 to start a company, I just didn't have the cost vs risk balance right to do it any early. That said, when I worked for others, I worked as hard as if it was my own company, so the fact I am now the beneficiary of my own hard work is the best part.

Recommended Articles

Fundraising is interesting! I have raised over 2m SGD for Musiio, but I didn't know anything about VC funding before I started Musiio, but as I said if you are naturally curious you can figure out just about anything.

Why did you found Musiio? What are the problems it is trying to solve?

14+ years ago when I worked at HMV I put the new CDs out for sale on Monday morning. A quiet week would be two new singles, a busy week maybe five or six, and that was it. The grand sum of new Western music was 10-30 singles a month -- a very manageable amount.

Today companies like Spotify report northward of 40k new songs A DAY added to their platform. This new volume is not manageable. You can't manually listening, tag or playlist in that volume. So we use AI to do those tasks.

What were the initial challenges in developing the tech for Musiio?

Our AI is right at the cutting edge of what is possible using technology, that makes it hard to develop and expensive to run. I compare it to websites. 10 years ago you needed a developer and a large budget to get a website. These days you need 100 bucks and a Wix account and anyone can do it. AI is still in the early stages, so it's high-tech and generally quite custom.

Why is the issue of mistaken or wrong metadata such a difficult and intractable issue in the music business?

It's an issue if the artist name is wrong for reconciliation of monies owed to artists, but at Musiio we are addressing a different issue. If 40k songs are uploaded to Spotify everyday (and every other streaming service globally) and it's not physically possible to manually listen to all of them, how does anyone know what they have? What genres the tracks are? What mood? If we can automate adding “rich metadata” (not just the artist and title but descriptors) to the audio, then songs become infinitely more searchable and playlistable.

AI is a buzzword that has been used ad-nasueum (for investors, press etc), often without it being applied in an impactful way. How is Musiio using AI in a real and meaningful way?

Good question! AI freaks a lot of people out too. I often say, our AI isn't sentient, it doesn't talk, it doesn't make life or death decisions, it just sorts music. That said, at Musiio AI and ML can be used somewhat interchangeably (this will make some developers pretty mad!)

ML, Machine Learning is the process by which you teach a computer to replicate a part of human behavior, in our case "hear a song" "assign a genre" if you teach it genres it can recognize the patterns the next time it sees a track. AI is the name used for anything given the appearance of replicating human behavior; this can be anything from walking/talking humanoid robots to having a computer identify a genre (which humans are also good at, just not in the same volume as AI). The former is definitely more “sci-fi.” I use AI rather than ML because it's a broader description. At Musiio we are using raw audio and training neural networks, we are the real deal. We are not just pulling feeds from open APIs and claiming data science is AI.

What do you see the main applications for Musiio being in the next five years?

Catalogues are growing, every day more and more music is released so the challenge we address only grows. Our goal is to make more music discoverable, to give the user the experience they want via smart tools and give the artist the opportunity to be found in the 40k a day.

Who are some of your clients at Musiio?

We work with Sync Companies like Audio Network and Epidemic Sound, Boutique Music House's such as Overcoast, gaming companies like Amanotes and sports performance companies like ClicknClear -- anyone with a large catalogue can benefit from our services.

Related Content