Jun 2022 – Jan 2025 · 2 yrs 7 mos · Bengaluru, Karnataka, India

• To remove content violating Illegal Regulated Commercial (IRC) policy of LinkedIn, built and deployed BERT based classification model
• Built and deployed Siamese Network based text similarity model to identify the off-topic posts in LinkedIn groups which was also adopted for AI-assisted Pending Posts Management for Group Admins
• Used Linkedin domain-adapted BERT model to build a Kindness Reminder classifier that triggers a warning when a user posts an offensive or unkind comment on LinkedIn
• To catch Money Fraud content, built a BERT based model which out-performed existing 1D-CNN based classifier
• Built a model to catch promotional content on Linkedin using author features, engagement features and post-text.

Oct 2019 – Apr 2022 · 2 yrs 6 mos

• Built a chatbot using Rasa that will assist users in buying beauty and personal care products
• Using Seq2Seq with attention and Transformer architectures built models that will generate the product description in natural language by utilizing product attributes.
• Created Image Caption Generation model that will give captioning suggestions to fashion influencers while posting on Myntra Studio.

Nov 2017 – Oct 2019 · 1 yr 11 mos

• Attribute Extraction from Fashion Images:
• Built individual multi-class classification model for each of the attributes wherein I incorporated differential learning rates across various layers that helped in improving average accuracy
• Created the branched progressive model for attribute extraction. With this approach only one model can extract all the attributes which is easier to deploy and training and inference time also get reduced by half
• Experimented with Residual Attention Networks model as well and created feature map visualization to ensure if correct portion of an image is getting attention by the network while doing classification
• Fashion Brands Similarity:
• Using descriptions, price and other attributes of products belonging to various fashion brands, created Word2Vec based model that will give the similarity between fashion brands at BAG (Brand-Article Type-Gender) level.

Mar 2019 – Aug 2020 · 1 yr 5 mos · Bengaluru, Karnataka, India · Remote

• Mentored more than 200 students, providing support and guidance for their capstone AI projects.

Mar 2019 – Jul 2020 · 1 yr 4 mos · Bengaluru, Karnataka, India · Remote

• Mentored 10+ students in Data Science and ML curriculum completion, also guided them in choosing and completing AI projects.
• Successfully helped students achieve their learning goals and gain practical experience in the field.

Aug 2016 – Nov 2017 · 1 yr 3 mos · Bengaluru, Karnataka, India

• Built a Lead Scoring Engine for a Leading Journal Publication Company of Europe
• Performed Customer Segmentation for a premier paperie and online retailer of USA
• Solved the Logistics Network Optimization problem of finding optimal locations of Regional Distribution Centers (RDCs) for a Leading Automobile Manufacturer of India
• Built a Recommendation Engine for a Leading video-on-demand Service Company using ensemble of Content-Based, Collaborative Filtering and Matrix Factorization algorithms

Jun 2014 – May 2016 · 1 yr 11 mos · Greater Bengaluru Area

• Optimized real time bidding strategies on major Ad-Exchanges for retargeting Ad-campaigns to maximize revenues and profit margins for the company and ROI for the clients
• Analyzed user data to understand click through patterns & buying behavior that helped increase customer engagement & conversions
• Was part of the team involved in creation of Programmatic Bidding Engine which calculated the worth of an Ad Impression in real time in RTB platform

Aug 2012 – Apr 2014 · 1 yr 8 mos · Gurugram, Haryana, India

May 2011 – Jul 2011 · 2 mos · Ghent Metropolitan Area

• Measurement and Simulation of Organic Light Emitting Diode Materials
• Designed and built the integrating sphere setup to measure transmittance and absorbance spectrum of various scattering layer samples used in manufacturing of Organic Light Emitting Diodes (OLEDs)
• Measured spectrum of an OLED sample and analysed the changes in spectrum with change in input current
• Simulated the External Quantum Efficiency of Green and White OLED taking into account radiative and non-radiative decay of exciton and assessed the importance of non-radiative decay
• Optimized the OLED light output as a function of electron transporting layer thickness