# Year 2 @ ServiceRocket

I think how a company compensates an employee is an indicator of how much the company values an employee. This is my attempt to list down noteworthy things I have done over the year, to help you review my compensation. Thanks!

## Moo Initiatives
- Built Moo migration pipeline from the ground up. Took the initiative to almost fully automate the process so that anyone could do the migration. We only have to trigger a Jenkins job and everything else will be taken care of!
- Aggressively pushed for migrating instances, this would set us free from freakin Contegix. To-date we have [migrated 45%](https://docs.google.com/spreadsheets/d/1LMmdTLx6zZmYlZulgTKUaDQmBclgWWn3vx5qX8T4cA8/edit#gid=2030153085) of learndot customers from Contegix to Moo.
- Did extensive work on migrating Mulesoft and did my best to document the findings. We also did a couple of dry-runs and could already successfully migrate them over to Moo. Checkout [the epic](https://trainingrocket.atlassian.net/browse/LD-6184) in all its glory.
- Compiled a comprehensive list of suggestions from my personal notes on Moo as my concluding remarks. Though, I wish I implemented some of them and fully took advantage of the autonomy I had during my last few weeks in the Reliability squad. Context switching was especially difficult during the late stages of FB03. [Workplace post](https://servicerocket.workplace.com/groups/svrkt.s611/permalink/2840172402730974/)
- End-to-end test for the entire Moo platform. In an effort to increase reliability of Moo platform, we’ve implemented automated end-to-end tests to continuously validate end user operations (like provisioning new instances, installing SSL certs) always work as expected. [Workplace post](https://servicerocket.workplace.com/groups/learndot.kul.nx01/permalink/2969427383103039/)
- Had to come up with quite an elegant way of specifying Jenkins branch, so that child jobs get receive the `branch` as their input parameter, i.e. nested testing. [Workplace post explaining the method](https://servicerocket.workplace.com/groups/eng.tooling/permalink/2591712157604916/)
- [PlayDays] Created a pipeline that builds any TR branch and provisions a brand new learndot instance under 10 minutes! One of my biggest playday projects as I have been working on this on and off for at least 6 months. Received a lots of positive feedback, checkout the [Workplace post](https://servicerocket.workplace.com/groups/eng.play/permalink/948691908828237/)
- [PlayDay] Zero downtime learndot deployment. This is a big deal, while we lose all states, we can achieve this very easily. Checkout this [demo](https://servicerocket.workplace.com/groups/learndot.kul.reliability.xp/permalink/2778156752251196/).
- Debugged and fix [Cloudera-Boomi issue](https://servicerocket.workplace.com/groups/learndot.kul.reliability.xp/permalink/2831871943546343/). Yes, I'm very proud of this. It was one hell of a "bug".
- GitOps is now widely adopted xoxo [1](https://servicerocket.workplace.com/notes/learndot-reliability-experience-squad/towards-gitops-managing-helm-state-via-git/2477263192340555/) [2](https://servicerocket.workplace.com/notes/learndot-reliability-experience-squad/gitops-declarative-way-of-doing-operations-with-vcs/2447391718661036/
)

## FB03
- I was given the freedom to work on this 2 days a week, and another 2 days was spent working on Moo.
- My role in FB03 was focused primarily on Machine Learning aspects of the project. And also mentoring the team with architectural decisions & machine learning aspects.
- Did extensive work on researching and improving the performance of our BERT baseline model [Detailed information on this Workplace post](https://servicerocket.workplace.com/groups/472316023397887/permalink/524236001539222/) you can check the teams comment there too. My most proud, breakneck work I've done so far at SR. Thanks to @Sanjev for going out of his way to get me this opportunity and for trusting and giving me the freedom! I'm really grateful for this and the "ML days" before that <3
- Since I only get to work on this for 2 days, I did not had the time to write a lot of posts. Couple of posts I did write:
- Help the team to onboard with Kubernetes and familiarize with our hackweek project code (Related post)[https://servicerocket.workplace.com/groups/472316023397887/permalink/473729873256502/].
- The team struggled with the long deployment process with Kubernetes on GCP for quite a while. I researched and setup Skaffold for our project which allows instant deployments [Workplace post](https://servicerocket.workplace.com/groups/472316023397887/permalink/528669381095884/)

## ML Initiatives
- Main aim is to grow machine learning capability in SR.
- Gave an introduction to Kaggle competitions to encourage participation and shared how semantic search works. [Recording](https://servicerocket.workplace.com/groups/2355691661378069/permalink/2532731533674080/)
- Encouraged others in ML SIG to participate in Kaggle competition, lead by example by also participating in two Kaggle competitions to show ML isn't hard to get into. Happy to see others working on Kaggle machine learning problems during their play days. [Workplace posts](https://servicerocket.workplace.com/groups/eng.play/search?query=kaggle)
- Sharing my learnings from improving the accuracy of the Semantic Search model for FB03. I'm still drafting two posts, one to share my experience competing on Kaggle & working with TPUs; and another one explaining BERT architecture holistically.
- Currently working with a couple of engineers on improving the performance of a model that identifies COVID-19 cases using X-rays. [Workplace post & discussion](https://servicerocket.workplace.com/groups/2355691661378069/permalink/2666511783629387/)

## GCP Initiatives
- Spearheaded the initiative to set up ServiceRocket's GCP account by working with Guoy, YC, Jessie Miller (Learndot's ex-GM) and Pei Yu (our GCP Account Manager).
[Checkout the group](https://servicerocket.workplace.com/groups/svrkt.s611/) for detailed timeline and updates on this
- Managed to get us $7,000 USD in credits for both Green and Orange - we set up separate billing accounts from the beginning
- We got the credits just in time before Hackweek. We used the credits and [won the Hackweek project](https://servicerocket.workplace.com/groups/eng.hack.weeks/permalink/879025619148249/) with [Semantic Search](https://servicerocket.workplace.com/groups/eng.hack.weeks/permalink/857272894656855/)! This will become one of the key aspects of million dollar project, FB03.
- Did a research on recent pricing changes in GKE and how we should approach multi cluster architectures. [GCP Workplace group](https://servicerocket.workplace.com/groups/svrkt.s611/permalink/2840172402730974/)
- Currently working on cutting down costs for `sr-ml-green` by cleaning up resources provisioned by FB03 squad.

## Community Engagements
- GDG Talk. Gave [a talk on AutoML](https://www.gdgkl.dev/) that discussed how to create high accuracy custom models and technical details on how AutoML works behind the scenes. [Slides](https://docs.google.com/presentation/d/1ebFNem53WUrY65QPeEWRamsf9LCjaIKhg3w1y01JEtk/edit?usp=sharing). Based on the event organizer's survey, this was one of top 5 talks! "60% find your talk incredibly helpful (with votes 4 & 5)" "52% find you are an awesome speaker (with votes 4 & 5)"
- Participated in Hackathons teaming up with fellow rocketeers: [Grab Challenge](https://servicerocket.workplace.com/groups/eng.play/permalink/879809779049784/), [Axiata's Urbanlytics](https://github.com/kornesh/urbanlytics-datathon) and [LawTech Hackathon](https://www.facebook.com/lawtechmy/media_set/?set=a.424474438259288) (We were top 5 finalists)
- I actively participate in meetups (mostly Tensorflow, Python, DevOps and GCP; sometimes KLJS). Besides drawing experiences from the community & soaking up good practices in the industry, I also promote ServiceRocket in these circles as a good place to work and often share that we're hiring (especially during "Who's Hiring" sessions).

Attending meetups gives me exposure to real world experiences of fellow practitioners; and an opportunity to critically discuss with the thought lords of the field on various perspectives, conundrums and trade offs that exist in the real world. Plus people skills, thought leadership and stuff. Competing in Hackathons and Kaggle competitions helps me sharpen up my hands-on technical skills. Both of which directly helps me improve the quality of craft that I produce at SR.

## Company level initiative
- We were using NPS as a de facto metric for learndot [you can check out the all hands] even though we only had really really small sample size. Wrote a post after extensively researching the weakness of NPS; to raise awareness of the limitations of metrics that we use and avoid overemphasizing their role in decision making.
- Received comments from a lot of the executives, including our CEO Robert Castaneda, Learndot GM Bill Cushard and our Customer Experience Lead Meliha Bilge. Checkout [the Workplace post](https://servicerocket.workplace.com/groups/learndot/permalink/2442893119283771/). And had a lot of offline discussion with others as well.