Ashok Krishna

Data Scientist, currently working as the head of the Data Science team at Unmetric with 3+ years of experience in identifying, building and managing various Machine Learning projects which are key and unique services offered by Unmetric compared to its competitors.

At Unmetric I ideate, implement and manage various Data Science projects, manage a team of 2 and report directly to the Head of Product & Co-Founder.

Projects

• Image Analytics

Object Detection – Used TensorFlow’s pre-trained Object Detection Model (Faster R-CNN Inception Resnet) for out-of-the-box inference. 

     Dominant Colour Detection – Developed an algorithm which outputs the Dominant Colours of an Image using KMeans and a colour difference metric (Delta E CIE 2000).

• Automated Campaign Finder

Developed a methodology to automate a slow and unscalable process of finding Social Media Campaigns run by brands. This job was done previously by a team of 5 for 1200 social media profiles.

• Reach & Impressions

Modelled the complex performance of a Facebook Post to predict its Reach. Improved the model performance from an Error Rate of 37%(version 1) to 9%(version 3).

• Promoted Post Detection 

Built a Random Forest model with a 96% accuracy (F1 score of 0.92) which predicts if a Facebook Post was promoted or organic using the Interactions Data (Likes, Comments, Shares) of a post. 

• Text Analytics

     Trending Hashtags – Developed an algorithm which takes brand tweets and mentioned tweets to find trending hashtags brands are jumping on daily using Chi-Squared Test.

Word Cloud – Developed a sophisticated algorithm which extracts important topics, phrases, hashtags and emojis. 

• Data Analysis

Collected user/client interaction data from various sources (Mixpanel, Database, Google Analytics etc…), automated analysis pipelines, constructed KPIs, and designed and automated dashboards to measure a product/feature success.