SARATH R NAIR

Senior AI Engineer

CONTACT

adoqm2@r.postjobfree.com

949-***-****

Bengaluru, KARNATAKA

LinkedIn

Github

LIBRARIES DEVELOPED

tf-transformers: State of the art faster

deep learning library for NLP in

Tensorflow 2

https://legacyai-org.medium.com/tf-

transformers-f7722536ba61

https://github.com/legacyai/tf-

transformers/

EDUCATION

Masters

Computational Engineering and

Networking

Amrita Vishwa Vidyapeetham

India

2012 - 2014

GPA: 7.5

AREAS OF INTEREST

Applied Statistics Linear

Algebra Computer Vision

Open Source Contribution

SUMMARY

Working as AI Research Engineer, with 7+ years of experience in Data Science, Deep learning and Natural language Processing focusing on making use of AI to help and uplift the business, society, and economy in a positive way. Always exploring new techniques, algorithm, and engineering, that will justify the business growth along with personal learning experience

Would love to be a part of diverse team to share and learn new ideas and perspectives, make use of data to automate and assist the existing systems to benefit the business.

SKILLS

Machine Learning

· Deep Learning + Neural Networks

· Natural Language Processing

· Search + Information Retrieval

· TensorFlow + JAX + PyTorch

· Python + Elasticsearch + MongoDB + Docker

· HIVE + pySPARK (basics) + EMR

· Basic Statistics + Image Processing

· Language Modelling using GPT2, BERT, DeBERTA, t5, LSTM, Attention, Transformers.

tf-transformers

Creator

· tf-transformers is a library developed in TensorFlow 2, for NLP based on Transformer Architectures.

· Developed smallest (parameters + computation) models with a GLUE score of (81.0).

· Faster than existing frameworks for TensorFlow 2 by 80 %. WORK EXPERIENCE

Senior AI Engineer

Fidelity Investments

2019 - current

· Working in Applied Research lab, developing and analyzing data across multiple business units, to derive insights and values.

· Developed, tested and deployed custom models, ranging from Logistic Regression models to state-of-the-art Language Models like t5, Roberta etc.

· Worked with cross functional business units to develop Custom Similarity models with primary focus on Search efficacy, improves the retrieval capability of the system from 63 % to 90 %.

· Research ranging from Zero Shot classification to Auto Regressive modelling.

· Used Uncertainty in Neural Networks to assess and quantify the risk associated with Benefit calculations to mitigate overflow, which not only have monetary impact as well as savings company's reputation.

· Focusing on optimizing and improving the performance of smaller and sustainable task specific models.

· Current Projects related to Structured Information Extraction and Contextual based Search, to leverage multiple data sources. NLP/ML Consultant

Consultant

Sep 2017 - Dec 2018

· Mentored a small team to develop multiple NLP models and deployed as micro services, as part of end to end Automation Assistance System.

· Involved in Research and Development to develop and guide Projects.

· Developed and deployed intent and slot filling models, to assist various stages of Chat boat assistance.

· Profanity filtering, spelling correction, toxic identification for Voice Based Assistance Systems.

· Containerizing application as micro services using docker and deploying at scale in Kubernetes (part of the team ).

· Developed and deployed TFlite models for edge device predictions and host-based modeling.

Data Scientist

MediaIQ Digital

May 2016 - Jul 2017

· Worked in Research and Development, with prime focus on analyzing Terabytes of data to draw useful patterns from user logs to target audiences for Digital Advertisement.

· Analyze competitor brands, using some brand specific data and keyword extraction, to recommend and fill the gaps in existing campaigns.

· Location-specific advertisement branding, re-targeting and prospecting using, co-ordinates and fixed region.

· Mapping coordinates to geohash and to postcode for efficient targeting.

· Ad targeting using URL + keyword classification using unsupervised Word Embeddings and deployed it in real time using efficient caching and clustering.

· Basic Exploratory Data Analysis using pySpark and HIVE on EMR and EC2 instances.

Data Scientist

May 2014 - Apr 2016

· Worked with the team on natural language Processing, to address different business problems in the HR and Recruiting space ranging from Information extraction to Search.

· Skill Recommendation using word embeddings (word2vec) from unstructured texts like Resume, Job descriptions etc.

· Information Extraction and Entity Recognition from resumes and job descriptions.

· Resumes/ Job description classification based on industry using Logistic Regression, RNN using TensorFlow

· Data Wrangling + Indexing using Python, MongoDB and Elasticsearch to build scalable search systems.

· Projects: Resume Parser, Skill to Resume Matching System, Skill Recommendation System, Skill to Resume Search System

Contact this candidate