Annual Report FY21-22: Building a chatbot

Product & Engineering

Carousell: Chatbot

Carousell uses Zendesk to power their support help desk which contained written articles and a chatbot. Their chatbot didn't allow us to configure the chat journeys and their recommendations systems wasn't tuned for us, and it was costing Carousell around Rs. 7-8L/month. A senior engineer had done some initial exploration and written a few RFCs on how to architect the system. The work was split in backend and frontend and I was responsible to take the RFC and implement it for the backend. We had a separate engineer to setup parts of frontend since I wasn't well versed with frontend. Carousell has open-sourced Orion framework which it uses internally to power all the microservices and I was responsible for creating, owning and maintaining a new microservice. I had been given very well defined work with objective metrics for work well done and I got all the help I needed. This seemed a perfect way to get started as someone fresh out of college.

My manager and a senior understood the importance of learning how to slice up the work and after I did some explorations, we all sat together to slice up the tasks, create JIRA tickets, and schedule them. They believed in giving me as much freedom as required but intervening when something was taking more time than expected and that's how I learned to time-box issues so that they don't end up dragging me. Since our team didn't have a frontend engineer, we had asked for help from another team, and he managed to setup the barebones and did a handover before the backend was ready. This had a lucky coincidence since I was looking to get my hands dirty with frontend.

We had planned for two phases, where in the first phase, we would release it only in one language (English) and see how it's working out before choosing to release for multiple languages. After the backend was done, I integrated it with the frontend, did testing, put it behind feature flags for a controlled roll out and finally released it. I was so overjoyed once we started small roll outs and saw it working properly. We then did a full roll out that same work and it was getting some serious amount of users. This was the first piece of software that I had put in production which was accessed by so many users.

Then came time for Phase II, the DB was designed in a way to accommodate for multiple languages because this requirement was known at the time of designing, so we just had to wait for the translations to arrive so that we could get started. After the translations came, only a month of work was required in backend and frontend to release it in production. Again, we did a controlled roll out and after a month of a full roll out, it had around 50k users a day. My software had never reached this wide an audience earlier, and I was overjoyed that it worked properly. In the full year after release, we had only one downtime, and it worked correctly without any maintenance long after I was gone.

Carousell Support Chatbot

Carousell's tech team is quite mature and a lot of things like CI/CD, logging, infra, etc were already setup with very Release for multiple languages good practices which helped me in learning the ropes.


I have often been asked this question as to why I invest time in managing my equity portfolio myself considering that my capital base isn't enough to justify the time spent on this to which I have always replied, "It's the irrational bet I want to take on myself", but here is what I never told any of them, "If you don't take that irrational bet on yourself which can change your personal growth trajectory forever, then no one else is going to bet on you."

Here's the thing about irrational bets on yourself:

  1. You look like a fool doing it, and you learn to not be bothered by it
  2. It's almost always mispriced and the upside is huge and if planned well the downside is limited
  3. It's best to align the bet in a direction where one is effortlessly good at
  4. It takes a very long time to materialise which is why patience, innate curiosity, and love for the craft are crucial to survive, otherwise, it becomes too easy to give up or be burned out

12 years ago, I took an irrational bet to install linux, and it took a decade for the bet to materialise. Managing my own equity portfolio is another irrational bet I want to take on myself and think I am well prepared to 'pay the toll'.

In 2018, I started keeping good track of my finances (initially using ledger and hledger, now I just use Sheets/Excel with Dropbox paper). I like to keep a track of Balance Sheet so that I can understand my portfolio and make bets appropriately. I often take leverage (capped) to hold positions and pay interest on it. I made a 5 year plan of saving a minimum of let's say x amount and maximum of 2x amount. The minimum amount was the threshold I kept to qualify myself to start a startup and get married. I had budgeted out emergency fund, marriage fund, and startup fund so that I can be my own investor. These savings are supposed to be the cushion which let me enjoy life as well as pay for my irrational bets. I am close to reaching the minimum amount level, which has given me immense confidence about my direction.

If you ever find something you are effortlessly good at, don't let go of it, nurture it instead. It may look irrational to others, learn to live with it.

A few people reached out to me saying they enjoyed reading these reports and I am glad to know people are enjoying it. Such words of encouragement keeps me motivated to publish more of these in future.