Akshat Gupta

Lead ML Engineer

Bengaluru, Karnataka, India6 yrs 9 mos experience

Most Likely To SwitchHighly Stable

Key Highlights

Expert in Deep Learning and Computer Vision.
Proven track record in scaling AI solutions.
Innovative content generation using Generative AI.

Stackforce AI infers this person is a Deep Learning and Computer Vision expert in B2C applications.

Contact

Skills

Core Skills

Machine LearningDeep LearningData Science

Other Skills

Adobe PhotoshopAzureBig DataC++Computer VisionData AnalyticsFPGAGenerative AIHTMLImage ProcessingJavaKerasLeadershipManagementMicrosoft Word

About

I am an experienced Deep Learning practitioner and researcher who combines expertise in computer vision technologies to provide smart solutions to solve real-life problems. - Extensive experience in Tensorflow and OpenCV to build solutions either from scratch or incorporate from existing approaches. - In-depth mathematical knowledge of existing ML and DL techniques to understand and combine them. - Like to replicate the results to understand and learn the ins and outs of the approaches and techniques. - Loves to brainstorm and always open to learning something new. - Fuels energy from an open, collaborative environment. Keywords: Computer Vision, Machine Learning, Deep Learning, TensorFlow, OpenCV, Keras, Scikit-Learn, Neural Networks

Experience

Glance

3 roles

Staff Machine Learning Engineer

Promoted

Jan 2025 – Present · 1 yr 2 mos

- Working on Glance AI, creating digital avatars using selfies. Productionizing SOTA stable diffusion and LLM models to production, serving 300k users and scaling across India, US

Machine LearningDeep LearningTensorFlowComputer Vision

MLE III

Jul 2023 – May 2025 · 1 yr 10 mos

Working on content innovation (using Generative AI to extract short clips/ trailers) to serve 50M users

Generative AIMachine Learning

Data Scientist II

Jul 2021 – Jul 2023 · 2 yrs

Worked on increasing the product image quality. This led to a decrease in load time from 3-4 seconds to 0.5 seconds and 5x in size reduction.
Implemented Multi Arm Bandit (Thomson Sampling) for the content recommendation for new users.
This led to a 40% increase in new user engagement.
Deployed ML model for sending notifications to users for live streams based on affinity towards a particular live streamer. This led to a 30% increase in the conversion rate
Architected a real-time deep learning-based recommendation system for Roposo: a short video, and
shopping app. Increased the engagement metrics by 3x

Machine LearningDeep LearningRecommendation Systems

Société générale

2 roles

Software Engineer - Data Science

Promoted

Jul 2019 – Jul 2021 · 2 yrs · Greater Bengaluru Area

Using Data lake and Big Data technologies to historize the financial deals data to perform analytics .Developing Rest APIs on Azure cloud to fetch data from data lake and widgets to display on the dashboard.

Data ScienceBig DataAzure

Machine Learning Engineer

May 2018 – Jul 2018 · 2 mos · Bangalore

Built a Convolutional Neural Network with two fully connected layers to classify 4 different kinds of Societe Generale signatures using Keras.
Used Image Processing to detect signatures in a binary image by extracting features by iteratively calculating the center of mass and the angles subtended to it by the 4 corners using OpenCV library.
Used Template Matching algorithm to obtain differences between the incoming and outgoing documents by dividing them into grids and checking the outgoing grids with the respective neighbouring grids of the incoming document.

KerasOpenCVImage ProcessingMachine Learning

The university of tokyo

Research Internship

May 2017 – Jul 2017 · 2 mos · Japan

Implemented computation engine for Convolutional Neural Network based on the architecture which takes in 7 tilled input arrays from CPU for images which are fanned out to 2 kernels and 7 input weights from the DRAM of FPGA and calculates the weighted sum.
This engine was implemented on Xilinx Vertex -7 chip using both MaxCompiler and Verilog softwares.
The clock frequency for the computation engine was around 103MHz in MaxCompiler as compared to 106MHz in Verilog

VerilogFPGA