Vinayak Goyal 20D070088
Electrical Engineering Dual Degree ([Link]. + [Link].)
Indian Institute of Technology Bombay Gender: Male
Specialization: Communication & Signal Processing DOB: 24/11/2002
Examination University Institute Year CPI / %
Graduation IIT Bombay IIT Bombay 2025 8.54
Intermediate CBSE Ramsheth Thakur Public School,Kharghar 2020 94.20%
Matriculation ICSE Ryan International School, Kharghar 2018 93.66%
Pursuing Dual-Minors in Computer Science and Engineering and AI & Data Science
Scholastic Achievements
• Ranked in the top quartile of the Simon Marais Mathematics Competition participants (2022,2023)
• Achieved an All India Rank 349 in IIT JEE-Advanced amongst 200,000+ shortlisted candidates (2020)
• Achieved an All India Rank 744 in JEE-Main amongst 1.5 million+ candidates across India (2020)
• Awarded the prestigious KVPY fellowship by the Government of India with All India Rank 270 (2018)
• Among National Top 1% in National Standard Examination qualified for INChO, INPhO, & INAO (2019)
Publication
• “Cross-Domain Evaluation of Few-Shot Classification Models: Natural Images vs. Histopatho-
logical Images”- Ardhendu Sekhar, Amit Sethi, Vinayak Goyal, et al. (3 others). Pre-print submitted to
Springer’s CCIS series for inclusion in the best papers book from BIOIMAGING 2024
Internships
Dispersion Trading | Quantitative Research Intern | [Link] (May-Aug 2024 )
Received Letter of Recommendation from mentor for exemplary performance during the Internship
• Developed multiple Implied Volatility spread signals between pairs of highly correlated Index Options
and between Index Options and their respective components and traded on spreads of straddle prices
• Estimated IV spread by summation of correlated normals, Butterworth low pass filtering, and EMA
• Designed and refined multiple at-the-money, delta-neutral, and market-neutral strategies to hedge indices
• Implemented strategies like capturing Options mispricings and lead-lag signal between IV and premiums
• Final IV spread signal tested on six month index-index pair data had Sharpe 9.2 and 6.31% mean profit/trade
Phone Camera AR Latency Reduction | Software R&D Intern | Samsung Research Institute (May-July 2023 )
• Tested methods to reduce end-to-end latency for augmented reality in upcoming Samsung mobile devices
• Implemented client-edge processing between two mobiles, using the stronger device’s GPU for weaker one
• Created a complete architecture for the client side of the client-edge processing system using Java
Investigate Linguistic Features for ERC | Research Intern | Programme DesCartes (May-July 2022 )
• Collaborative research project of CNRS, NUS, and A*STAR aspiring to develop hybrid AI for a smart city
• Aimed to incorporate linguistic knowledge in DL models for Emotional Recognition in Conversations(ERC)
• Proposed and tested two hypotheses on effects of linguistic coordination on emotions of speakers in a dialogue
• Analyzed and modified the DailyDialog dataset to identify specific features and make model based on them
Key Projects & Competitions
Image Super Resolution using Shearlets | Master’s Thesis | Prof. Amit Sethi (June 2024 - Present )
• Designed a multi-feature untrainable pipeline for Shearlet transform from scratch with NumPy and PyTorch
• Developing an architecture to extract meaningful features from early layers to reduce model complexity
• Achieved PSNR 20.99 by using shearlet transform block as feature extractor followed by Resnet architecture
Reinforcement Learning | Course Project | Prof. Shivaram Kalyankrishnan (July-Nov 2023)
• Implemented multiple regret-minimizing algorithms such as UCB, KL-UCB and Thompson Sampling
in various multi-armed bandit instances and compared their performance with theoretical expectations
• Applied Value Iteration, Policy Iteration, and Linear Programming methods to compute optimal policy
• Developed an RL algorithm to play optimal action in billiards with directional noise depending on speed
Few-Shot Image Classification and Segmentation | RnD Project | Prof. Amit Sethi (Aug 2023 - Jan 2024)
Paper submitted in CCIS Journal for review
• Explored Interactive Few-Shot Learning by understanding, and implementing PFENet, PSPNet, and
Generalized FSSeg for classification and segmentation and tested on PASCAL and COCO medical datasets
• Tested SimpleShot and LaplacianShot on medical datasets trained on Natural vs. Histopathology images
Generating Image by Denoise Diffusion Models | Course Project | Prof. Sunita Sarawagi (Jan- May 2023 )
• Aimed to modify images by identifying an ‘h-space’ and a corresponding additive vector in a diffusion model
• Successfully completed the project with the ability to adjust “extent of smiling” through h-space scaling
Incorporating Shearlets in WaveMix | RnD Project | Prof. Amit Sethi (Jan- May 2024)
•Designed Wavelet and Shearlet based processing blocks inspired from WaveMix to improve its architecture
•Achieved top accuracy 87.05% using above blocks as compared to 83% without them on CIFAR-10 dataset
Chest X-Ray Disease Detection by Deep Learning | Course Project | Prof. Kshitij Jadhav (July-Nov 2023 )
• Implemented Federated Learning by partitioning client-server data and effectively preventing data leakage
• Tested on NIH X-ray dataset with 100k+ images from 30k+ patients, for multi-label classification of 14 diseases
• Improved mean AUC in ROC curve from 0.62 to 0.75 by making a noise removal layer using Haar Wavelets
Data Fusion for Satellite Image Classification | RnD Project | Prof. Biplab Banerjee (Jan-June 2024)
• Designed a UNet based & an attention based architectures to extract LiDAR and HSI features for segmentation
• Modified FusAtNet with DeepLab-V3 and hyperspectral attention for improved spectral feature emphasis
• Achieved 85.64% accuracy in pixel-wise classification on the Houston dataset containing 15 classes
Fingerprint and Retina Identification and Verification | Course Project | [Link] (Aug-Dec 2023 )
• Engineered and programmed a Wavelet-based and a Shearlet-based feature extractor for retina and
fingerprint images using adaptive thresholding, sobel and gabor operator to use with Siamese network
• Achieved maximum identification accuracy of 85.71% in the Siamese network with only 3.6 Mn parameters
Domain Adaptation (DA) for Signal Compression | RnD Project | Prof. Satish Mulleti (Jan- May 2023 )
• Used DA to develop a class specific compression-decompression model to improve the compression ratio
• Created and tested autoencoder models for domain adaptation, transfer learning and few-shot learning
• Achieved denoising in MNIST dataset along with compression ratio 1/196 and reconstruction loss <-20dB
Worldquant Quantitative Research Challenge | Team Member | Inter-IIT Tech Meet 12.0 (Oct-Dec 2023 )
• Represented IIT Bombay at Inter IIT Tech Meet 12.0, excelling in the Brain Quantitative Analyst Challenge
• Secured 4th rank among 21 IITs by exploring 100+ datasets, 10000+ simulations, and presenting 35 alphas
• Generated alphas by automating API using genetic algorithms, technical and field correlation analysis
LSTM-based Stock Trading System | Course Project | Prof. Amit Sethi (Aug-Nov 2023 )
• Achieved MSE loss 0.0002 in price prediction using sixty-minute window with more than one correlated stocks
• Created trading logic using price prediction by LSTM models trained on prices of groups of correlated stocks
• Developed a flexible LSTM model for price prediction, integrating a dataloader and trading decision logic
NLPlay With Transformers | Summer of Code | Web and Coding Club (May-July 2021 )
• Built positive/negative sentiment classifiers using FFNN, RNN, LSTM, GRU, and transformers
• Achieved best accuracies 93.5% and 93.91% on LSTM and GRU respectively by testing on IMDB dataset
• Undestood Attention and applied BERT, RobERTa, XLNet & DistilBERT for sentiment classification
Positions of Responsibility
Teaching Assistant | EE 353: Introduction to Data Science and Machine Learning (July 2024-Present)
•Member of a small cohort of TAs responsible for assessing scripts and assignments for a batch of 300+ students
•Part of team to support instructor through the course for grading, conducting quizzes and re-evaluations
Design Engineer | Motor Controller (MC) Design | IITB Racing Team (Dec 2020 - March 2023 )
• Part of Motor Controller Design subsystem in Electrical division of the IITB Racing team for 2.5 years
• Redesigned the schematics and optimized the board layout for MC’s control card with Eagle Autodesk
• Mentored trainees and Junior Design Engineers, leading update meetings to ensure clarity on project tasks
Key Courses & Technical Skills
Programming Languages Python, C++, C, VHDL, Java, Arduino
Data Science libraries pandas, NumPy, PyTorch, TensorFlow, Keras, OpenCV, SciPy, Matplotlib, NLTK
Electrical Engineering Probability and Random Processes, Wavelets, Image Processing, Cryptography,
Markov Chains and Queuing Systems, Online Learning and Optimisation
Mathematics Calculus-I & II, Linear Algebra, Complex Analysis, Differential Equations-I & II
Computer Science Introduction to Machine Learning, Advanced Machine Learning, Reinforce-
ment Learning, Principles of Data and System Security
Extracurriculars
• Participated in IMC Prosperity-2 & Tower Research Capital’s annual Limestone Data Challenge
• Represented Hostel 6 in the Swimming and Triathlon General Championships against 7 other hostels
• Among top 6 teams of Institute Techincal Summer project to develop a Robotic Glove prototype using Arduino
• Successfully completed one year of professional training under the National Sports Organization
