I am currently on the job market seeking ML engineering & research opportunities starting Dec 2023!

I'm a second-year masters student in Computer Science at The University of Texas at Austin. I'm advised by Prof. Chandrajit Bajaj and affiliated with the CVC Lab @ Oden Institute. My research is centered around Reinforcement Learning (RL), Bayesian methods, sequential decision-making ML and their applications. Currently, my work is motivated by problems in the domain of medical image diagnosis.

Before coming to UT Austin, I worked on heteroscedastic uncertainty quantification in segmentation models and VAEs with Prof. Chandrajit Bajaj. Prior to that, I researched & engineered generative language models at the Web Science Lab @ International Institute for information Technology, Bangalore (IIIT-B) under Prof. Srinath Srinivasa.

I graduated with my B.Tech. in Computer Science from Mahindra University, where I worked with Prof. Arya Bhattacharya on gradient-free optimization algorithms and Prof. Bruhadeshwar Bezawada on algebriac methods for symbolic execution, respectively. In my free time, I enjoy running, playing tennis (and squash), and traveling!

News

Aug 2022: New preprint on optimal sampling for tensor decomposition out on Arxiv now!

Nov 2021: Gave a talk on Multi-Agent RL for efficient Rank-Ordered search @ Modern Inverse Problems Annual Meeting, RWTH Aachen University!

Jun 2021: Will be joining UT Austin MSc.CS program in Fall 2021!

Mar 2021: Two papers on training neural networks using differential evolution accepted @ IEEE CEC 2021!

Preprints

Learning Generative Embeddings using an Optimal Subsampling Policy for Tensor Sketching.

Chandrajit Bajaj, Taemin Heo, Rochan Avlur

Under Review, 2022.

[arXiv][code soon]

Publications

Comparative Performances of Neural Networks of Variant Architectures Trained with Backpropagation and Differential Evolution.

Zakaria Oussalem, Rochan Avlur, Jhanavi Malagavalli, Arya Bhattacharya

IEEE Congress on Evolutionary Computation (CEC), 2021.

[Paper]

Training Convolutional Neural Networks with Differential Evolution using Concurrent Task Apportioning.

Rochan Avlur, Zakaria Oussalem, Arya Bhattacharya

IEEE Congress on Evolutionary Computation (CEC), 2021.

[Paper]

Lossless video compression using Bayesian Networks and Entropy Coding.

Rochan Avlur, Chandrasekar Vaidyanathan

IEEE Region 10 Symposium, 2019.

[Paper]

Projects

Investigating Causal Overhypothesis.

Investigated representations in parametric latent variable models (LVMs) & proposed weighted-mixture scheme to induce compositionality.

[Paper][Code]

Generalization via Adaptive Planning in Model-based RL.

Investigated role of trajectory evaluation for open-loop model-predictive planning in model-based RL.

[Paper][Code]

Controlling Estimation Bias in Q-learning.

Proposed a multi-arm bandit setup for online adaptive control of estimation bias in Q-learning based on Lan et al.

[Paper][Code]

Learning Node Characteristics for Reliable Leader Election.

Proposed a multi-agent multi-arm bandit algorithm for consensus in asynchronous distributed systems; reduced node failure up to 20%.

[Paper][Code]

Representation Damage in Pruned Multi-modal Transformers.

Investigated influence pruning large multi-modal models (CLIP) have on latent representations by analyzing multi-modal neuron activations & performance on downstream tasks.

[Paper]