Leading the Engineering and Cybersecurity effort at Writesonic

- Lead multiple projects/teams while working hands-on full stack.- Implemented Assembly Line Platform (ALP): The ALP stock platform takes away the garbage project management work and allows the managers to focus on coaching and deep dives. It is a worker-oriented system that provides a factory mode of operation, focused on increasing the productivity of teams. We onboarded ALL the assembly lines in Trilogy to this platform. My role here included full stack development, leading a team of 7 people and making process decisions with the stakeholders.- Implemented ARM Devspaces (Provision Dev Environments with AWS Graviton2 instances): Trilogy has an in-house product called Devspaces which provisions dev environments in a matter of seconds. It uses x86 AWS instances for compute. I lead the effort to add the support for AWS Graviton2 instances (ARM based) for compute. My role here included architecture planning, hands-on development and leading a team of 4 people.

- ML Factory: Any product should be able to export their data to the factory and get ML driven insights within 1-2 weeks. ML factory aims to streamline the data analysis lifecycle through automating Ingestions, Transformations, EDA. It supports efficiently running parallel experiments on processed data and logging metrics/plots out of the box. It addresses ML problems related to predict, personalize and forecast. My role here included leading the team, and implementing ML models, and designing an orchestrator to help data-scientists run and log their ML experiments, get best performing model deployed, and run real-time or batch inference on test data.- KB articles assessment automation: Lead a team of 4 developers to automate the assessment of knowledge base articles (using NLP) to help Trilogy's support teams function faster.- Autofix production source code: This product is aimed at finding insights in Java codebases and automatically fixing them for the Trilogy portfolio of products, helping save millions of dollars per year. My role here included architecting and designing the fixer algorithms while leading a team of 9 developers. Worked on autofixes related to duplicate code, long methods, NPEs, Resource Leaks.

- Implemented a mini-Java calculus to facilitate theorem proving in the theorem solver called Coq. Designed a formal mathematical proof for (dis)proving equivalence between two code snippets for the refactoring problem of long methods and duplicate code fixes.- Lead a team of 3 developers and implemented a system to generate unit-test templates for the mutation paths in Java code, with a mutation coverage of more than 90%.- Implemented a system to identify probable points of failures inside Java codebases in a product release. This entire system is built serverless using AWS services like Lambda, Fargate, QuickSight dashboard, and gives insights across four verticals at a package level: test coverage, static analysis findings, hotspot analysis and impact analysis.

- Designed and implemented a rule-based system to automate the tedious and time-consuming process of doing code-reviews. Integrated static-analysis tools like PMD into the system. The tool finds violations across 450 Java and C# rules with an accuracy of above 99%. This tool was widely adopted within the org and aided in automating 90% of work done by Trilogy's code-review team.- Implemented a vectorized model of source code which would allow training ML models on code. Collaborated with source{d} to evaluate and apply transfer learning on SOTA algorithms.

Trilogy university is a 3-month Bootcamp modelled on the US Navy SEAL training. As part of the program, I mentored 20+ new hires on software development, took tech-sessions on Databases, Docker, Deployments, Kubernetes, various AWS services, and managed three projects to drive towards success.

As a Udacity Project Reviewer, I review student project submissions, striving to give actionable and helpful feedback. The areas in which I provide project reviews are Artificial Intelligence, Machine Learning and Deep Learning. I also help in beta testing of the projects in Deep Learning Nanodegree and Artificial Intelligence Nanodegree.

Software Intern - Trust and Safety Team, Fraud Detection Division.Identifying probable fraud sellers at Flipkart's platform and automating the process to shop for products corresponding to such sellers.