brainsteam.co.uk/brainsteam/content/posts/2024/11/end-of-an-era.md at 83a561aa659c0fce04bc1a8a34edb679024e42d4

12 KiB

Raw Blame History

date

description

draft

mp-syndicate-to

preview

Early Days and First Revenue

We started Filament as a 'boutique' AI & ML consultancy in the summer of 2016. We set out to solve data science challenges for companies small and large using whatever the best approach might be. The founding team were a bunch of experienced folks who had built and sold a digital agency (websites and stuff) previously. I initially joined on a part time basis while working on my PhD at Warwick Uni. I brought the software engineering and ML know-how; they brought their business acumen and rolodexes.

The first few months were fun but also scary. We spent a lot of time doing business development, experimenting with different offerings and trying to find a market fit. It was the height of Pokemon Go mania so one of the proofs-of-concept we worked on was an augmented reality money-saving app (think something like Shopmium or JamDoughnut) where you ran around a physical store to find 'deals' and interact with them on your phone.

It was also the height of the IoT craze. With some outside help from a very talented maker friend, we built a pro-bono interactive Christmas tree for Alder Hey Children's Hospital. The tree had "smart" lights and a little screen in its base. When you mentioned the tree on twitter or put money into a physical coin receiver (a donation to the hospital trust) the lights would flash and a little message would pop up on the screen. We had a bit of a nightmare with couriers damaging parts of the tree and spent the night before we presented it to the hospital fixing it up in our AirBnB. With that project we learned an early lesson. Maybe hardware wasn't for us...

{{

}}

Even after a few short weeks, we made inroads with some really massive companies thanks to the packed out rolodexes that the other co-founders brought along. However, even with a strong network, getting a completely unheard-of 5 person company through procurement at a multinational is nearly impossible. I remember sitting around in our somewhat grotty Soho Studio just before Xmas 2016 with about 12 weeks of runway left (we were completely bootstrapped and running on fumes). We were due to hear about a big deal with a German multi-national that would save our bacon for another 6 months if it came in. The phone rang and the deal landed the day of our christmas party. We collectively breathed sighs of relief and bought cheap champagne from a Piccadilly off-license. We drank it out of plastic dixie cups.

Winning Clients and Cool Projects

After the initial struggle, we found a successful business model selling data strategy and proof of concept projects. I'd do a lot of the data science work myself, exploratory analysis and modelling and write a personal report showcasing what we found. In some cases this led to repeat work or we'd be asked to help with hardening up the prototypes for production. In others, the company's data would yield no insights or they would take the product internally.

I thoroughly enjoyed these projects and the variety of companies, people, problems and technologies that they exposed me to. My favourite project was helping a commercial aerospace logistics company to build a model that predicted which departure gates we should park planes at, taking into account delays, weather signals etc. We also did a pilot study on fast food self-service kiosks for a major international chain with great results. We didn't hear back after we presented our results but shortly afterwards they acquired a recommendations platform company. We (jokingly) like to take credit for that.

During Covid we had the opportunity to help out on Project OASIS which was a collaboration between the UK Government, Academia and a couple of private companies aiming to knit together data collected from covid symptom tracker apps. The project was reliant on voluntarily submitted symptom reports and did not have any contact tracing capabilities. Eventually we were completely (and rightly) blown out of the water by the official NHS covid tracking app. However, I am still really proud and honoured to have been part of that team and experience.

{{

}}

While we kept the lights on with consulting projects, we were also building out Filament Engine. Engine was an MLOps platform. This was long before the term MLOps had broken into the zeitgeist. We showed it to a few clients, but at the time it wasn't clear to them why they would need such a platform and we tried to do way too much with a skeleton crew. We were trying to build Weights&Biases AND HuggingFace hub AND Argilla in a single platform with a team of 8 people and no investment before the market was ready for any of those things.

Eventually we wound down the Filament Engine project pivoting to EBM, a chatbot management system which we span off as its own business in early 2023. We also started some pilots with banks and private equity companies that would eventually be sewn together and become Syfter, modern Filament's main product line.

Building a Team I'm Very Proud Of

Over the years it has been my absolute pleasure and privilege to build out a tech team made up of incredibly talented and easy-to-work-with people. No mean feat for someone who left their previous role with only very limited managerial experience. I just knew how I liked to be treated based on my experience of working in other software teams and I tried to foster a culture of trust, autonomy and curiosity. The co-founding team trusted me to make the right calls (although in the early days they'd often all be involved in interviews). I'm sure there was an element of luck involved too...

The first hire I made was a bright graduate who I had befriended after working together at IBM when he was a student intern. He ended up staying and growing with the company and now remains a really key senior engineer on the Syfter platform. We couldn't grow quickly, especially in those early days whilst we waited for news that our first project was going to land. As we started to win consultancy projects, we began to expand the team organically.

In 2017 we hired a really great technical people-manager via our network at IBM and he helped me set up an office space in the Southampton/Solent area. The founders' mandate was for us to build "a developers' paradise". I don't know if we quite got there but we did have a pretty nice little office space which we initially filled with Ikea furniture after a dash into Southampton one afternoon. I've written about our solent office spaces previously. From there, we built a little technical community on the south coast, many of whom are still with the company today.

We got lucky with our ability to attract experienced seniors via our network and we also pulled in talented students and recent graduates, many of whom stuck around for a long time or remain with Filament today. We also benefited immensely from the UK Government-funded Knowledge Transfer Partnership which allowed us to hire masters and PhD-level data scientists and get academic supervision from top tier universities including Warwick and Essex. Working with both was a pleasure and led to me working with some truly brilliant award-winning academics and associates. I'd highly recommend hitting up Essex Uni if you ever consider starting an AI research project. I hope I will have the pleasure of working with them and the lovely folks at Innovate UK again in the future.

{{

}}

Of course people don't stay in jobs for ever. Some of the folks I've mentored or managed have ended up going on to work for really prestigious organisations and while I'm always sad to see them go I love to see them thrive in a new environment. I'm glad that I got to work with some of these people during their meteoric rise. I've always enjoyed lecturing but my time at Filament helped me to realise that 1:1 mentoring and teaching is even more rewarding and I've made many friends for life in this role.

So Why Leave Now?

So why now? Why move on? Filament is still growing and selling Syfter to an ever-expanding user-base.

The focus of the company has naturally changed a lot since I joined. Whilst I truly am proud of the hard work, blood, sweat and tears that have gone into the Filament Syfter platform, a big part of me yearns for the variety that used to come with general consultancy. I'm looking forward to having new opportunities to explore more varied problems and technologies once again. Likewise, Filament needs and deserves a technical leader who is passionate about the current mission rather than one who is nostalgic for the old days.

The team I've been lucky enough to assemble around me at Filament is mature and fully functioning. I'm not a single point of failure any more. In the early days of the company I'd end up opening my emails or slack after time away and because I was lead dev, QA data scientist and SRE, there was always something wrong. I remember (not fondly because us Brits like our PTO) taking a call from one of the devs on a mini-golf course in Weymouth during a family holiday where he wanted me to talk him through OAuth security flows. However those days are thankfully gone now and after a recent 2 week vacation I returned to only 150 new emails and a handful of non-urgent slack messages. I'm so happy and proud to have a self-sufficient team and I know that they will be absolutely fine without me.

After 8 years, I've become far too comfortable. I'm probably only 1/3 of the way through my career and I don't want to get stuck in my ways just yet! It's time to shake things up again and try something new!

{{< figure alt="A photo of a group of adults (the filament team) sat around a long table eating food out of takeaway cartons" src="https://media.jamesravey.me/i/0b982fb0-1ba6-410c-a2d1-9862615d8317.jpg" title="The filament team are a sociable bunch and I'm going to miss team lunches like this one very much"> }}

12 KiB Raw Blame History

Early Days and First Revenue

Winning Clients and Cool Projects

Building a Team I'm Very Proud Of

So Why Leave Now?

12 KiB

Raw Blame History