Divyam Arora

AI Researcher

Hyderabad, Telangana, India8 yrs 11 mos experience

Most Likely To SwitchHighly Stable

Key Highlights

Expert in building customer-facing AI models.
Proven track record in improving data efficiency.
Strong background in machine learning and NLP.

Stackforce AI infers this person is a Data Scientist with expertise in Fintech and AI-driven solutions.

Contact

divyamarora910@gmail.com LinkedIn

Skills

Core Skills

Machine LearningNlp

Other Skills

A/B TestingAWSAnomaly DetectionBig DataCasual InferenceCommunicationCreative Problem SolvingCreative ThinkingDBMSData ScienceData VisualizationData WranglingDeep LearningExperimentationExploratory Data Analysis

About

Data Scientist with more than half a decade of experience in data science domain. Currently building customer facing AI models. Seeking to increase data efficiency of Microsoft AI in Edge Browser R&D Team for data Science and machine learning problems. Skilled in Machine Learning, Data Science, Data Wrangling, Big Data( Hive, Teradata), Predictive Modelling, Forecasting, Sequence Models (Deep learning), Data Structures, Algorithms & creative thinking. Specialities: Python, Building solutions using Machine learning, Deep learning(sequence models) and Engineering, Exploratory Data Analysis, Data Wrangling.Data Structures and Algorithms. Currently working on Big data Machine learning and Sequence Models. My interest lies in building AI solutions and building customer facing models and similar work as a data scientist/ML scientist/Research Scientist.

Experience

8 yrs 11 mos

Total Experience

2 yrs 11 mos

Average Tenure

4 yrs 2 mos

Current Experience

Microsoft

Applied Scientist 2

Apr 2022 – Present · 4 yrs 2 mos · Hyderabad

Edge Growth ML Team, Microsoft AI Division
Currently building Reinforcement learning system to automatically learn user behavior and open edge browser automatically at his peak usage intention.
Deployed and Built ML System for predicting next url a user will browse to improve user browsing rate and nudge him into engaging more on edge browser. Stay tuned for new future update of web browsing on edge.
(Exploring BARTs , Small language models like Microsoft Phi 2, BERTs )
Researched and built multi variate BERT including temporal and other dimensions
Improved search on Edge Add-on store( edge browser extension store) using RAGs based contextual search and deployed on Azure which work alongside lexical search
Worked on improving bing chat suggested questions at the edge browser launch start using NLP and clustering from urls of a client to improve user interest in the chat questions
Built a measuring method for decision making on browser interactions of user with casual inferencing
skills used: ML design,Machine Learning, NLP, Statistics, A/B Testing, Experimentation, Casual Inference

Machine LearningNLPStatisticsA/B TestingExperimentationCasual Inference

Paypal

2 roles

ML Scientist 2

Promoted

Dec 2021 – Apr 2022 · 4 mos · Chennai, Tamil Nadu, India

Worked on Location intelligence first of it's kind at PayPal using Open Street Maps as research project. Used AWS services and Open Street Maps.
Built credit BNPL ( paypal pay later ) ML underwriting model for PayPal international markets like UK, Europe, integrated location intelligence into this as well.

Location IntelligenceAWSOpen Street MapsMachine Learning

Associate Data Scientist

Aug 2020 – Dec 2021 · 1 yr 4 mos · Chennai, Tamil Nadu, India

Global Data Science- Credit & Sales, India
Built customer facing AI model like credit promotion (propensity) marketing model for UK PayPal, In FY 21 and 22 if you lived in UK and were using paypal app, if you received a marketing email saying that you are pre approved for paypal credit, my model was the reason you were sent that email after checking your credit worthiness.

Highradius

4 roles

Data Science Engineer - 1

Jul 2019 – Aug 2020 · 1 yr 1 mo

Created Automatic extraction of PDF claim document using Spacy (deep learning) and Nearest
Neighbor Algorithm and engineering concepts(inclination points, angular view, regex tagging).
Achieved 85% accuracy and reduce engineering time by 75%.
Published a whitepaper on the company's portal named "Automated Extraction of Information from Claim Documents".
Lead a trigger word detection for company’s AI assistant feature engineering using Spectrogram,
CNN and model using LSTM with 88% capture rate for trigger word.

Deep LearningSpacyNearest Neighbor AlgorithmMachine Learning

Associate Data Science Engineer - 2

Promoted

Aug 2018 – Jun 2019 · 10 mos

Developed a predictive model for the identification of valid/invalid deductions as Anomaly Detection problem for an invoice.
Achieved more than 60% automation rate with less than 0.003% automation risk and 85% capture rate. This significantly reduced the number of transactions they have to manually review since 60% are automated, freeing up their time for high-value work i.e looking at 85% captured invalid deductions first which increases the ROI.
Built Generalized Machine Learning Pipeline to solve the deduction validity problem to reduce the deployment time by 80%.
Presented multiple data analysis on deductions while recommending product features, core business & technical metric goals with PMs.
> Published a white paper in company's portal on "Nearest Generative Samples for model explainability" for better explainability of deductions models.
> Published a white paper in the company's portal on "Evolving ML model to mitigate self-induced data exclusion bias in Deductions model" for removing the biased created by suggested actions from the models.

Predictive ModelingAnomaly DetectionMachine Learning

Associate Data Science Engineer -1

May 2017 – Jul 2018 · 1 yr 2 mos

Building end to end machine learning and deep learning solution with the main goal to increase ROI and automate the Account Receivables and Treasury Products as much as possible. Starting from selecting and getting the data sources from clients, cleaning, manipulating it and then analyzing it, making predictive models or deep learning models out of it and then deploying it in python web services.