Tasha Pais
tdp74 at rutgers dot edu

I'm a fourth year Computer Science and Cognitive Science student at Rutgers University. I'm fortunate to be advised by Kostas Bekris and to be a member of the Physics-aware Research for Autonomous Computational Systems Lab (PracSys) at Rutgers. Here, I work on generative 3D research with Shiyang Lu.

I think robotics is a quest to understand the human mind. In addition to common robotics problems like learning, perception and reasoning, I've worked on neurolinguistics (language production in Broca's area) and mechanistic interpretability (vision transformers).

My personal philosophy has changed a lot. Currently it most closely aligns with pyrrhonian skepticism, empty individualism, and absurdism. As we get closer to a post-scarcity world, I've thought a lot about what motivates me. For now, it's accelerating our understanding of cognition and more broadly building deeptech companies.

CV / Twitter / Github

Updates

Research

Revolving-View 3D Object Completion Using Text-Conditioned Inpainting

Tasha Pais, Shiyang Lu, Kostas Bekris
In progress
Github

Active Exploration: Scaling Up Tactile Sensing Algorithm for Category Level Classification

Tasha Pais*, Nikhilesh Belulkar*, Jingxi Xu
July 2023
Github

Hair Dryer Behind Table
Extending Semantic Abstraction of 2D VLMs for Efficient Search of Hidden Objects

Tasha Pais*, Nikhilesh Belulkar*, Huy Ha
December 2022
PDF   •   Video   •   Github

* indicates equal contribution, active exploration video by REAL

Projects


I'm currently working on these ideas:

Computational Robotics

Generating control sequences within a robot’s limits and integrating forward with added noise, implementations of particle filter, kalman filter, and extended kalman filter for localization
December 2023
Report   •   Github

RRT and PRM implementations for a rigid body with interpolation along a straight line in the c-space
November 2023
Report   •   Github

2d collision checking for convex polygons using ray casting, collision-free navigation using minkowski sums to compute configuration spaces up to SE2 dimensionality
October 2023
Report   •   Github

Machine Learning

Jacobian Chain Rule and Linear Algebra Refresher   •   Everything I Know Written As Tiny As Possible
Bayes risk, gaussian based generative classifiers, implementing EM for GMM, comparing evaluation on diagonal and full covariance matrices
December 2023
Proofs   •   Colab

Softmax over negative margins of the ensemble to derive adaboost, decision trees using gini impurities
November 2023
Proofs   •   Colab

Kernelized SVM in dual and primal form, kernel trick, vector representation of input text, stochastic subgradient descent using pegasos
November 2023
Proofs   •   Colab   •   Handwritten notes

Logistic regression models and equivalent softmax models, l2 regularization and cross entropy loss, negative log likelihood in forward pass, classification on MNIST
October 2023
Proofs   •   Colab

Least squares estimation, maximum likelihood estimation, asymmetric squared loss
September 2023
Proofs   •   Colab

Operating Systems

Custom malloc for virtual address allocation; two-level page table for 32-bit address translation; direct-mapped TLB for efficient address translation caching; bit manipulation for efficient memory tracking; designed a 4-level page table for 64-bit addressing; ensured thread safety and compatibility with various page sizes, benchmarked using matrix multiplication.
November 2023
Report   •   Github

Thread creation, yielding, exiting, joining, and synchronization using mutexes; scheduling policies including Pre-emptive Shortest Job First (PSJF) and Multi-Level Feedback Queue (MLFQ); thread context management through makecontext, swapcontext, and ucontext APIs
October 2023
Report   •   Github

Blockchain

Quadratic voting application deployed on polygon, developed in Hardhat and Next.js, smart contract tests written in solidity and javascript, inspired by Vitalik's blog post advocating for nonlinear cost functions
December 2022
Github

Consulting

In partnership with the UN, presented at 2021 general assembly; building low-cost, cloud computing institutes in Dharavi, Mumbai; refer to slides 28 and 29 for break-even calculations with net output and capital cushions
May 2021
Deck

click for more projects