Each year our faculty members invest their time, effort & innovation in some great real-time projects.The projects are listed below. A student can apply for a max of 03 projects.
Registration will start from 1st March 2023. The last date to apply is 31st March 2023. The results will be up on the website by
15th 21st April 2023 & the internship will commence from 6th May 2023.
Process to apply for internship
- The details of the project openings are listed below.
- Click Here to Apply. (Closed Now)
- Login with your google account
- Fill the registration form. If you want hostel select the checkbox accordingly.
- For hostel booking details please click here
- To apply for any project just click the apply button
Students who get selected for the summer internship at IIIT-Delhi will be receiving a stipend of 5k per month
Summer Internship Projects 2023 at IIIT-Delhi
|S.No.||Faculty Name||Project Name||Project Detail||Duration||Status|
|1||Anubha Gupta||Benchmarking DL models for medical image segmentation||In this project, we will identify 5 datasets of medical imaging, identify 10 most recent DL state-of-the-art models, implement and benchmark these for object detection and segmentation.||3 Months||Physical|
|2||Anubha Gupta||Benchmarking Self-supervised classification models for tabular data||In this project, we will work with multiple datasets ( tabular medical and non-medical datasets) and benchmark self-supervised models as well as clustering models for unsupervised classification.||3 Months||Physical|
|3||Anubha Gupta||Benchmarking of Out-of-Distribution Detection DL methods||In this project, we will identify the recent state-of-the-art DL methods of OOD detection and benchmark their performance on 3-4 medical and non-medical datasets.||3 Months||Physical|
|4||Anuj Grover||Semiconductors & Environment||Course Material Creation on "Semiconductors and Environment"||10 Weeks||Virtual|
|5||Arjun Ray||Elucidating the druggable chemical landscape for lipid pathway using structural biology, genomic and ML approaches”||Utilizing pharmacogenomic approach, the project shall revolve around utilizing drug repurposing for genes that shall be reflective of population variations in lipid biosynthesis pathway. The project shall involve protein structure modeling, virtual screening and ML techniques.||6 Months||Physical|
|6||Bapi Chatterjee||Distributed and Federated Learning||To implement and working on a scalable Distributed/Federated Learning Framework||2 to 3 Months||Physical|
|7||Debarka Sengupta||Pseudo-time trajectory analysis of evolving cellular DNA||We will build stochastic methods to model how cellular DNA undergo alterations.||3 Months||Physical|
|8||Dr. Saket Anand, Dr. Sanjit Kaul, Dr. Manuj Mukherjee||Understanding autoencoders - Can they mimic speakers?||Autoencoders are a pair of neural networks, commonly referred to as the encoder and decoder, which work as follows. The encoder takes a data sample,
such as a speech sample, and encodes it to its latent space representation. The decoder takes the latent space representation and it reconstructs an
estimate of the data sample.
A recent work has shown evidence for the conjecture that autoencoders can mimic speakers. In more detail, an autoencoder is initially trained using voice samples from a single speaker. Next, the autoencoder is fed test samples from other speakers, and it was shown experimentally that the decoder is able to reconstruct the exact speech, but in the voice of the speaker from the training samples. We hypothesize that the encoder neural network is somehow capturing the semantic characteristics, i.e., the contents of the speech, in the latent space representation, while the decoder neural network is able to add the
speaker characteristics to the latent space representation during the reconstruction.
In this project, the student is expected to test out this hypothesis by designing the necessary experiments. If convincing evidence is found to support this hypothesis, the next goal will be to develop some theoretical understanding of the functioning of autoencoders using the insights gained from the experiments.
Necessary qualification: The student should be well versed in training neural networks using python. In addition, basic mathematical maturity, including some familiarity with probability and linear algebra, is also required.
|9||Manohar Kumar||Trustworthy and Explainable AI, Public Interest||There are two projects. In one the intern will be working on broad areas of Ethics of AI and in another in Political Philosophy on questions of public interest||2 Months||Physical|
|10||Ojaswa Sharma||AI-based detail enhancement of 3D medical imaging data||This project will explore and investigate AI based approaches to enhance details in 3D medical imaging data such as CT and MRI.||3 Months||Physical|
|11||Ojaswa Sharma||AI-based segmentation of 3D medical imaging data||This project is about designing a segmentation neural network for MRI and CT medical volumes. We will utilise a high quality ground truth data for fine tuning the network to achieve high accuracy.||3 Months||Physical|
|12||Piyus Kedia||Tools for automated reasoning||This project aims to build tools to help students learn the internals of SAT and SMT solvers.||2 Months||Physical|
|13||praveen priyadarshi||Urban Public Spaces Post Liberalisation: A Study of Delhi-NCR||The objective of this research is to explore and understand the nature of urban public spaces in contemporary Delhi-NCR. The project is premised on the idea that the nature of ‘publicness’ is a constitutive element of the ways in which urban politics functions on an everyday basis.
|14||Ranjitha Prasad||Unsupervised federated learning and anomaly detection||Federated learning (FL) is proving to be one of the most promising paradigms for leveraging distributed resources for collaborative training of a machine learning model, while protecting the privacy of individuals' data. To leverage the enormous unlabeled data on distributed edge devices, we aim to use the FL paradigm in unsupervised tasks, and in particular, we address the problem of anomaly detection in decentralized settings.||2 Months||Physical|
|15||Richa Gupta||Accessible space design with AR for visual and hearing impaired||Several public spaces can be made more accessible and convenient to use for persons with visual and hearing impairments with location based AR tagging. This project will entail field study of spaces such as metro stations and use of AR for improving the accessibility for visual and hearing impaired.||2 Months||Physical|
|16||Rinku Shah||In-network computation using FPGA NICs||This project involves development of network applications using HLS (High level synthesis) programming for FPGAs. The pre-requisites for this project include foundational knowledge in the domain of computer networks and experience with HLS programming for FPGAs.||2 Months||Physical|
|17||Sanat K Biswas||Orbit Computation of Resident Space Objects for Space Situational Awareness||There are more than 20000 man-made objects of more than 10 cm in size floating around in near-earth space which pose collision threats to functional satellites. Predicting collision probability from these space objects is crucial from the national security perspective as well as for the protection of public and private space assets of Indian origin. The outcome of this project will directly support the Indian Space sector (valued at INR 51,334.85 crore) by providing an operationally flexible, scalable, transparent and indigenous collision probability solution.||2 Months||Physical|
|18||Sneh Saurabh||ASIC Design using open-source EDA Tools (Yosys, OpenSTA, OpenRoad)||In this project, ASIC design using open-source EDA tools will be undertaken. The emphasis will not only be on the design but also on understanding the internal mechanics of these tools. The intern will be expected to understand the C++ code of these EDA tools and make some minor modifications to the tool.
The required skillsets:
1. C++, data structure, algorithm
2. Working experience in UNIX
3. Digital Circuits (BTech 1st-2nd year level)
Desirable: ASIC Design Flow
|19||Tavpritesh Sethi||How well does AI “understand” genomes?||This project involves exploring and exploiting the features of transformer based AI models for biological and clinical decisions in the real world.||3 Months||Physical|
|20||Vinayak Abrol||Bandwidth extension (BWE) in automatic speech recognition||The aim is to extract features from speech that are informative irrespective of the sampling rate. One could think of using a generative network to upsample the audio. Complex rich DNN models can also be explored. Here, we are recovering missing frequency content- Bandwidth Extension in Audio. Alternatively, one can explore employing some sort of attention mechanism to directly extract informative features Or a mix of both approaches.
Ideally we want to implement this in a ASR pipeline using NVIDIA Reva platform. Some of the attention based ideas might come from Computer Vision domain. And of course there will be concepts from DSP and Speech processing.
|21||Vivek Bohara||Intelligent reflctive surfaces for Satellite communication for 6G cellular applications||The students would be doing literature survey on existing communication technologies for satellite communication. Further, they would be investigating on how deployment of intelligent reflective surfaces may improve the performance of existing technologies.||3 Months||Physical|
|22||Vivek Kumar||Parallel Execution of Quantum Computations||Design and implementation of C++-based parallel constructs to enable parallel
execution of a quantum kernel using a header only Quantum computing library (https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0208073#sec001)
|23||Vivek Kumar and Sanat K Biswas||C++-based Parallelization of Orbit Computation of Resident Space Objects for Space Situational Awareness||Parallelizing orbit computation of resident space objects for space situational awareness on multicore processors using C++ based task parallel programming model. Relevant background: a) https://ssl.iiitd.edu.in/project/space-situational-awareness/, and b) https://github.com/habanero-rice/hclib/blob/master/tutorial/hipc18/presentation/hipc2018_slides.pdf||2 Months||Physical|
|24||Vivek Kumar and Sumit Darak||Dynamic Task Parallelism on FPGAs||Implementing a work-stealing load balancing runtime for FPGAs. A paper describing on such implementation: Here||2 Months||Physical|
|25||Pankaj Jalote||Enveave is an environmental project being developed at IIIT Delhi (Front End Developer)||At Enveave, our purpose is to empower grassroots communities and individuals for environment action. We are developing a one-stop platform for India’s communities to come together and access services, resources and information to be successful in their environmental initiatives. The foundational tech platform for Enveave has been built and we are now in the deployment, implementation and testing phases. Click Here||3 Months||Physical|
|26||Pankaj Jalote||Enveave is an environmental project being developed at IIIT Delhi (Back End Developer)||At Enveave, our purpose is to empower grassroots communities and individuals for environment action. We are developing a one-stop platform for India’s communities to come together and access services, resources and information to be successful in their environmental initiatives. The foundational tech platform for Enveave has been built and we are now in the deployment, implementation and testing phases. Click Here||3 Months||Physical|
Last updated: 19-04-2023