Anamitra Singha

Welcome to my portfolio! I’m Anamitra, a Master’s student in Information Technology at IIT Delhi under Prof. Aaditeshwar Seth. My research centers around AI-driven solutions to tackle socio-economic challenges, and this portfolio showcases my work in deep generative modelling, machine learning, deep learning, reinforcement learning, and other areas in AI.

https://indiandude123.github.io/portfolio/

🧑‍🎓 About Me

I’m from West Bengal, India, with a B.Tech in Biotechnology from NIT Durgapur and professional experience as a Senior Analyst in backend engineering at Capgemini. My journey in AI began in my third year at NIT Durgapur, leading to an internship at IIT Patna under Prof. Sriparna Saha. In 2023, I joined IIT Delhi’s MS by Research program to deepen my knowledge in AI.

Education

M.S. by Research in Information Technology, IIT Delhi (2023 – Present)
B.Tech in Biotechnology, NIT Durgapur (2018 – 2022)

Experience

Senior Analyst, Backend Engineering at Capgemini

Internship

Research Intern at Indian Institute of Technology Patna

📝 Publications

Sentiment and Emotion-Aware Multi-Modal Complaint Identification (AAAI 2022) Publication The expression of displeasure on a consumer’s behalf towards an organization, product, or event is denoted via the speech act known as complaint. Customers typically post reviews on retail websites and various social media platforms about the products or services they purchase, and the reviews may include complaints about the products or services. Automatic detection of consumers’ complaints about items or services they buy can be critical for organizations and online merchants since they can use this insight to meet the customers’ requirements, including handling and addressing the complaints. Previous studies on Complaint Identification (CI) are limited to text. Images posted with the reviews can provide cues to identify complaints better, thus emphasizing the importance of incorporating multi-modal inputs into the process. Furthermore, the customer’s emotional state significantly impacts the complaint expression since emotions generally influence any speech act. As a result, the impact of emotion and sentiment on automatic complaint identification must also be investigated. One of the major contributions of this work is the creation of a new dataset- Complaint, Emotion, and Sentiment Annotated Multi-modal Amazon Reviews Dataset (CESAMARD), a collection of opinionated texts (reviews) and images of the products posted on the website of the retail giant Amazon. We present an attention-based multi-modal, adversarial multi-task deep neural network model for complaint detection to demonstrate the utility of the multi-modal dataset. Experimental results indicate that the multi-modality and multi-tasking complaint identification outperforms uni-modal and single-task variants. —

🔬 Projects & Case Studies

1. Conditional Deep Convolutional Generative Adversarial Network (cDCGAN) on Fashion-MNIST

Overview: Developed a Conditional Deep Convolutional Generative Adversarial Network (cDCGAN) in PyTorch to synthesize class-specific, realistic images resembling the Fashion-MNIST dataset.
Highlights
- Implemented both the generator and discriminator as deep convolutional neural networks (DCNNs), enabling high-fidelity image synthesis and improved training stability.
- Incorporated class-conditioning by integrating label information into both generator and discriminator, allowing targeted generation of specific Fashion-MNIST categories.
- Designed a flexible command-line interface supporting configurable training regimes, including variable generator/discriminator update steps and class-conditional sampling.
- Automated the end-to-end ML workflow: custom data loading with label management, model checkpointing, and visualization of class-conditional generated samples throughout training.
- Provided comprehensive project documentation and usage instructions to facilitate collaboration and open source contribution.

2. Hidden Markov Model for Sailboat Trajectory Simulation

Overview: Simulation of sailboat trajectories using Hidden Markov Models, simulating random wind effects and noisy sensor readings.
Highlights:
- Developed trajectory sampling, likelihood estimation, and Viterbi decoding algorithms.
- Used the Baum-Welch algorithm for parameter learning.
- Visualizations include sensor probabilities and KL divergence analysis.

3. Automated Micro-Suturing Analysis using Computer Vision

Overview: Computer vision pipeline for analyzing micro-sutures in medical imagery, automating precision-based metrics.
Key Features:
- Applied preprocessing techniques and edge detection for suture identification.
- Measured inter-suture spacing and alignment metrics for quality assessment.
- Automated CSV report generation for consistent evaluation.

Overview: Training agents in reinforcement learning for environments with discrete and continuous states.
Approach:
- Implemented Policy and Value Iteration for TreasureHunt environments.
- Built SARSA and Q-Learning agents for model-free learning.
- Applied DQNs with CNNs in LunarLander-v2 for optimizing exploration and stability.

🏆 Achievements

Rank 8 HackerEarth Machine Learning Challenge: Love in the Time of Screens
Top 4% Resolving Citizens’ Grievances HackerEarth Challenge
Ranked 126/3.3k Teams Amazon ML Challenge

📂 Skills

Programming: Python, C, SQL
Technologies: PyTorch, AWS, Django
Specializations: Machine Learning, Computer Vision, Reinforcement Learning, Data Science, Statistics

🌐 Connect

I’m always open to collaborative projects in AI, ML, and data science. If you’re interested in discussing a project, reach out!

Email: saheb123.singha@gmail.com
LinkedIn: linkedin.com/in/anamitrasingha

This portfolio is a continuous work in progress—thank you for visiting!