CV - Ankit

Work experience

Machine Learning Engineer Intern: June 2022 – June 2023

Sensity AI, Amsterdam, Netherlands
Developed a Forgery Detection Network achieving 5.2% higher test accuracy than baseline for detecting counterfeit ID cards using PyTorch classification models: ResNet, MobileNetV2, SiameseNet, EfficientNet, Dual Attention Network, and Visual Transformers (ViT). As a result enhancing Sensity’s ID Verification and Document Authentication services
Dataset Generation using Computer Vision methodologies for augmented training of our Forgery Detection Network
Contributed to Sensity’s Deepfake Detection API developing a generalized deepfake detection model for detection-in-the-wild

Remote Deep Learning Research Intern: April 2020 - August 2020

Pattern Recognition Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany
Nanoparticles occur in various environments as a consequence of man-made processes, which raises concerns about their impact on the environment and human health. To allow for proper risk assessment, a precise and statistically relevant analysis of particle characteristics (such as size, shape, and composition) is required and would greatly benefit from automated image analysis procedures. Collecting a real dataset of silicon dioxide nanoparticles is quite expensive. To overcome this issue, I worked on developing a Differentiable Trainable Renderer for Generative Adversarial Networks for the synthetic image rendering of silicon oxide nanoparticles under the guidance of Prof. Andreas Maier.
Developed a Differentiable Trainable Renderer for Generative Adversarial Networks using PyTorch for Synthetic Image Rendering of silicon dioxide nanoparticles which effectively reduced the costs of dataset collection through TEM imaging

Machine Learning Research Intern: May 2019 - July 2019

Visual Computing and Analytics Lab, Indian Institute of Technology B.H.U, Varanasi, India
Developed a Feature Extraction Algorithm as inputs to classification models like SVM, Random Forest, Naive Bayes, and XGBoost achieving 2% higher accuracy than baseline for breast cancer histopathological image classification

Projects

Guiding Text-to-Image Diffusion Model Towards Grounded Generation:

Report of MSc AI Computer Vision-2 course project: Repo Link
Achieved 4% higher mIoU while utilizing only 25% of the total dataset generated using Stable Diffusion compared to the baseline approach for open-vocabulary segmentation
Investigated the impact of prompt engineering and vector maths for CLIP for image segmentation

Google Ads Keyword Idea Generator and Scorer:

Demo
Created an LLM-powered app extracting keywords, generating new ideas through a LangChain agent, and fetching Google Ads metrics
Deployable via Docker, served on Chainlit, with Redis-cached metrics

Publications

[Re] Counterfactual Generative Networks

Ankit Ankit, Sameer Ambekar, Mark Alence, Baradwaj Varadharajan. "[Re] Counterfactual Generative Networks." ML Reproducibility Challenge 2021, 2021

SKDCGN: Source-free Knowledge Distillation of Counterfactual Generative Networks using cGANs

Sameer Ambekar, Matteo Tafuro, Ankit Ankit, Diego van der Mast, Mark Alence, and Christos Athanasiadis. "SKDCGN: Source-free Knowledge Distillation of Counterfactual Generative Networks using cGANs." ECCV 2022 Workshop VIPriors, 2022

Leadership

President, Office of International Relations, NIT Tiruchirappalli (September 2020 – May 2021)

Successfully, led a team of 25+ members, spearheading campus internationalization policies under the ambit of the administration body of the Institute

International Exchange Global Volunteer at AIESEC in Poland May (2018 – June 2018)

Worked on project ’Understand’ under the Sustainable Development Goal (SDG) Reduced Inequalities

Ankit

CV

Education