Leaving Skit, and Vernacular

When I joined the company, we were pivoting from building enterprise chatbots (the kinds you see on websites) to call center voicebots. I was still an academic at the core and had to spend a lot time just to learn working with real-world users and customers. But during this phase, I also spent a lot of energy on work so I was still able to produce quite a good amount of value for the company.

Most of our focus went in shipping speech recognition systems in Indian languages and, in parallel, going live with voicebots at scale serving real callers, a feat that feels easy to do now, but needed a lot from the few young and naive kids trying to build a product whose core and peripheral technologies both were completely new to them.

This is the same phase where I started taking managerial responsibilities and we made a more formal ML team where I did everything involved in an Engineer's, Researcher's, and Team's life cycle. We built solid processes to hire ML engineers and got some really great folks onboard. Almost everyone was extremely passionate about their core skill set and a lot of novel ideas got implemented quickly and provided immediate benefits. Overall I feel this phase had really great grounding for building a stellar brand for our ML and Tech team, something that was lost in the later years.

Unlike a few companies, COVID wasn't disastrous for us. Which is obvious if you think about what we do, which is automating human labour. While working remotely, we hired extensively and got a lot of things right. The company kept scaling up well.

This was the phase where I did not do much tech work but learnt a lot about people and, as a side effect, myself. Being in my home all day reduced the environmental spontaneity to near zero. As a result, no thought went away on its own, and needed proper escorting via thinking and reflection. Not that I usually go out much anyway, but change of scene was a strong tool for managing stickiness of thoughts which we didn't have in those years. This was both good and bad for me.

On the good side, lockdown gave me a lot of time with myself allowing me to go deep in any idea, as well as acted as an equalizer (all conversations happened at reduced modalities) which helped me preserve confidence in my skills even with a variety of people of all levels coming in at rapid pace.

I got a chance to think deeply about how I want to work, what it means to build teams, how to interact with people. I also played with a lot with novel ideas around leadership, management, ML, and technology in general.

After COVID, a lot changed. Not primarily because of direct effects like people coming back and seeing each other in office for the first time, but because the indirect effects around economy rendered some value faultlines to be more visible than before. Additionally lockdown allowed me (and others) to explore my values and philosophy of life more deeply and I was able to connect my inner conflicts more cleanly with how things were happening, say, at work. So even when a crisis of old gravity came in, you could see more visible differences in reactions. With the benefit of hindsight, I believe this was a phase where we should have done a lot of work to build cultural cohesion. This was also the phase during which we had a layoff, which was—in my opinion—managed well but learnings from the events that lead to it could've been handled better.

In any case, slowly a lot of passion towards making tech and product diluted. Part of this was because we had to do more scaling than building at this time. But this transition wasn't compensated well from other dimensions. And when the time came to wear our maker hat again, (at least) I wasn't able to find one in the locker.

While visibly different, the company was not in a bad state (financially, it was bouncing back well and kept doing better than before) in this phase. But I had been personally accumulating a sort of disenchantment that made me consider more seriously my life outside of Skit.

Overall while a lot of what we did could have been done in lesser time, I am happy about what we ended up doing. We were the first in market across many dimensions. Some of our tech work is in our blog here. A few open sourced projects are here on GitHub. I have compiled our research publications in my wiki here.

Additionally, as an early employee and someone interested in problems beyond just technology, I was involved in a lot of things that no one else would touch, including running initiatives across HR and marketing. I attempted many—and failed at many—org wide transformations and setting up core processes and standards. I also ran the initiative to rename the company from Vernacular to Skit¹.

Then there are things that I have regrets for. Not in the sense that they eat me up everyday, but more like "If I had another chance, I will work on …". A few important ones are listed here:

1. Our core product is a voicebot that converses with callers to solve their problem. But we are a B2B SaaS selling this software to businesses running call centers, so caller experience becomes a secondary metric. This makes it harder to optimize for but I believe we should have made more impact than what we did. I wrote some of my views around this here.
2. I would like to improve my decisiveness in general. This, at times, led me to not make significant progress in directions, technical and non-technical, that I had un-validated positive hunches about.
3. I have historically not given much value to marketing and PR. This is okay at a personal level, but at my position and in the competitive startup scene this led to some disadvantages for our team and company. While I tried to counter this bias, I could have done more and handled this smartly.

I am tempted to write a deeper retrospective but that probably should be done in a separate post. For now, I will move on to the next question.

Like any other company, there were a few business decisions that we took that made me uncomfortable. But working in debt collections with an automation tech (voicebots) was the hardest for me. Here is Vice's coverage on us and this domain in general if you want to understand the use case.

Beyond this specific decision I learnt that it's hard for a B2B SaaS to focus on end-users (not our direct clients) which then mutes out a lot of passion for the product since, while it interfaces with end-users, the evaluation framework is held by clients.

I am not an idealist in this regard and I can't believe that businesses are either black and white, with primarily capital seeking businesses being always black. I believe there are different components that play different roles in a system. But I also eventually realized that Skit's was not the role I wanted to play. If I want to indeed work on Spoken Conversational AI, I would rather work on user-sided tech² like spam deflection than something that makes money for large companies at the cost (mental and monetary) of the masses.

For all 6 years, I led the ML team and worked in a core leadership position. Working this closely at top gives you a good view on how the company operates and, importantly, will operate in the future. And this is a good opportunity to figure out whether a team will work for you. In my case, I realized that the leadership circle and their philosophy of working wouldn't gel with me for the long term. Usually this is where conflicts in core values and principles come in picture. For a young startup though, such conflicts could take time to show up since everyone is growing and the true values are not what's written down in the beginning, but what you see visible year on year, specially during crises.

I understand that details of such conflicts are very contextual and it's not important to talk about them since they risk misunderstanding. But I believe that acknowledging differences and conflicts as a reason of exit is a healthy part of growing up so I do not want to hide this factor.

I found that spending my effort and time in abstract leadership roles was not putting me in a state of flow day on day. I am primarily motivated by technical innovation and problem solving and the opportunity to dive deep. More broadly I believe that roles that allow me to create things of beauty using my skills are what make me happy. It's not that heading the ML function at Skit didn't allow this at all, but the added responsibilities as an early employee along with being in an early stage startup kept me less free for such 'wants'.

Throughout my stay, I learnt more about startups and the reasons they exist—which is primarily to make money—, but I stayed unmotivated by the prospect of making huge amount of money. This made me more prone to disconnection as the company took turns in the tides of market. While my peers could anchor themselves on the future prospects and take compromises here and there, I slowly realized that I was unable to do so.

But it would be a lie to say that I left only because of the above. I also left because I wanted to move on and do some other things. I am a big believer that free time is essential for truly novel outcomes, and I couldn't get any if I was working in the same way as before. Among other things I wanted to:

1. Study many topics at depth. The list includes intelligent systems, but also languages, evolution of life, complex systems, and more.
2. Study philosophy and come out with something that I can call my own 'philosophy of life'. This has become increasingly important for me.
3. Figure out areas that fit in my way of life and produce impact that I could be happier about.
4. Work on my body, mind, and cultivate other interests, including my tech hobbies beyond programming.

Working on these with any form of seriousness needs time (working and leisure) which I can't get in a regular full time employment. Adding to the fact that having free time right now, when my brain is active, is immensely more helpful, I decided to take a long break-for.