Hello, Srik here.

As a Senior Data Scientist in the Trust & Safety Team

As a Data Scientist in the Bidder Product Team

As an MTS ML/AI SDE at AMD R&D

I developed a system that automatically generates test plans for drivers using Recurrent Neural Networks (RNNs) and Word2Vec. This system also classifies the failures and identifies the root causes of the failures. This project saved 9+ FTE resources for multiple focus areas in ~1.5 years.

I developed an ensemble of Random Forests and Naive Bayes models to identify the severity of bugs. This project reduced the need for manual evaluations by 30%.

As Co-Founder and CEO

I developed a content-based recommender system for songs and tracks. I scraped over 10,000 songs and tracks, automated the annotation of track elements, and created over 100 playlists. This project decreased search time by 75% and increased user engagement by 50%.

I led a team of four in deploying models to estimate optimal web page journeys and predict customer retention. This helped in better targeting of customer segments and improved customer retention strategies.

As an Analyst in the Trust and Safety team

I led the efforts to reduce spam and abuse on the platform by building scalable solutions.

I built an in-house machine learning team and managed the delivery of 5 projects within 2 quarters.

I developed an ensemble model to identify gibberish content on web pages, which reduced the spam rate by 35% in a quarter.

I built a tool that automatically segments user feedback into actionable segments for different teams using natural language processing techniques and machine learning models. This tool reduced the time to action by 20% and improved the feedback response rate by 50%.

I trained 150+ employees across different locations on machine learning techniques to increase ML adoption.

As a Senior Analyst

I used Principal Component Analysis (PCA) to identify the causal factors that contribute to album ranking on the Billboard Charts for a multinational music conglomerate.

I developed a strategy that used clustering techniques to convert prepaid to postpaid accounts for an Indonesian telephone services provider. This project improved monthly conversions from 3.5% to 15% in 2 quarters.

As an Analyst

I trained 2,500 models using various segments of a 2 billion row dataset with 2,500 dimensions to detect credit card frauds for a global credit card company. This project reduced fraud by 40%.

I modeled a credit scoring system for a Middle Eastern country's Credit Bureau with an AUC of 0.75.

I prepared brand perceptual charts by scraping Twitter data for a multinational beverage brand to identify potential market gaps.