Shubham Jain

Data Engineer

San Mateo, California, United States7 yrs experience

Most Likely To SwitchAI Enabled

Key Highlights

Expert in Generative AI and Deep Learning.
Proven track record in risk modeling and fraud detection.
Strong foundation in Computer Science from top institutions.

Stackforce AI infers this person is a Fintech expert specializing in AI-driven payment solutions.

Contact

: shubhja@iitk.ac.in ; shubhamjain1310@gmail.com LinkedIn

Skills

Core Skills

PythonGenerative AiDeep LearningRisk ModelingComputer Vision

Other Skills

Apache PigApache SparkArtificial Intelligence (AI)CC#C++CSSCaffeData StructuresHadoopJavaScriptKerasLaTeXLarge Language Models (LLM)Latex

About

I am currently part of the Foundational GenAI team at VISA Research where we develop foundational methods for payment. I received an MS in Computer Science from UIUC and a Bachelor of Technology in Computer Science and Engineering from IIT Kanpur. I have interned with Prof. Yoshua Bengio and at NVIDIA and Adobe.

Experience

Visa

3 roles

Data Engineer - Sr. Consultant level

Jul 2024 – Present · 1 yr 8 mos

Leading efforts for Generative AI models for payment at VISA Research

PythonPySparkGenerative AI

Staff Software Engineer

Mar 2021 – Jul 2024 · 3 yrs 4 mos

Leading productionalizing of Deep learning model for real-time predictions at VISA Research

Senior Software Engineer

Jun 2019 – Feb 2021 · 1 yr 8 mos

Risk Modelling team in Visa Research
Trained deep learning models to identify fraudulent transactions and handle extreme class imbalance
Models use RNN based methods and rely strongly on embeddings of the merchants as well as the cardholders which are learned using the Word2 Vec bag-of-words algorithm
Working on utilizing Bi-Directional RNN based networks in scenarios where a previously approved transaction can be revoked using future transactions
Also working on CNN based methods to generate embeddings for merchants as well as cardholders as an alternative to a variant of Word2Vec that we currently use

Deep LearningRisk Modeling

Nvidia

Software Engineer Intern in Deep learning

May 2018 – Aug 2018 · 3 mos · San Francisco Bay Area

Facial Landmark Network :
Trained a network (in Theano) which takes a face and gives facial keypoints on eyes, mouth, nose, etc as output which is used to do head pose detection and eye open/close detection
Converted this network to Caffe and defined a custom layer in Caffe for the Softargmax layer.
Converted the Caffe network to TensorRT with fp16 which reduced the runtime by 50%
Integrated the TensorRT model in the team’s pipeline to be used with other features like face-detect,head-pose, eye open/close, etc

Adobe

Research Intern

May 2017 – Jul 2017 · 2 mos · Noida, Uttar Pradesh, India

Glimpsing Neural Network for Form Image Parsing:
Worked on models to parse a high-resolution form to detect various things such as widgets etc
The underlying idea was taken from the paper Show Attend and Tell which is an attention model
based on CNN - LSTM for captioning
Created the dataset using MNIST digits on a black background and predicted the bounded box around the digits in an orderly manner

Montreal institute of learning algorithms (mila)

Visting Student Researcher

May 2016 – Jul 2016 · 2 mos · Greater Montreal Metropolitan Area

Research Intern under Prof. Yoshua Bengio, MILA working on Sound Generation:
Used stacked GRU as the basic model and quantized each audio sample as an embedding and added weight normalization to the GRU layers
Applied transformations like pitch shifting, time-stretch on audio which gave a better generalization
Further added conditional information like different actors and categories of sound in the GRU
model to generate a larger variety of sounds and also added Recurrent Dropout and Recurrent Batch Normalization to our model to make it more robust