Sachin Grover Email & Phone Number

Tempe, Az, Us

Tempe, Az

San Francisco Bay Area

– Created an LLM agent using ChatGPT3.5 to translate natural language request to formal PDDL command and used a planner to create the plan in AI2Thor simulator.– Coded and evaluated 132M parameters language model with results comparable to ChatGPT2.– Created a carbon emission chat bot support using ChatGPT3.5 and ChatGPT4.0 to evaluate numerical accuracy of LLMs.– Worked on DARPA EMHAT, to design evaluation methods for digital twins and AI agent teaming based on LLMs. Developing the system for Pandemic, a multi-agent collaborative game.

San Francisco Bay Area

– Won, DARPA AIE, EMHAT, to design evaluation methods for digital twins and AI agent teamingbased on LLMs. Developing the system for Pandemic, a multi-agent collaborative game.– Worked on DARPA project SAIL-ON, to design and develop Open-world AI based agent and evaluated it on Polycraft (Minecraft based simulator), Cartpole++ (3D inverted pendulum) and Angry Birds game.– DARPA KMASS for Knowledge MAnagement at Speed and Scale to develop a dynamic knowledge base for 20,000 documents and design update methods to scale to millions of documents.– Part of industrial projects with Panasonic to evaluate smart factory development for their production plants

I am working in the area of Automated Task Planning under the guidance of Prof. Subbarao Kambhampati as a part of Yochan Lab. My ongoing projects include - Creating explicable policies in a stochastic scenario. Explicable in a Human Robot team means that the agent should perform tasks in a way that a human observer can understand them. Like in motion planning the trajectory of motion helps human understand the goal of the agent. To tackle this problem I am working on learning human preferences and make a preferential policy for an MDP. Although, one can argue preferences mean explicability, but it is not actually so. Explicability represents the human expectations (which is made from the humans understanding of agents capabilities) and preferences are personal choices of how a particular goal should be reached.I am also going through the literature for applications of planning techniques to Intelligent Tutoring System.

I continue to be a part of the Dragoon Lab. Currently, working on a new Intelligent Tutoring System - TopoMath (https://topomath.asu.edu). Implemented the logic to modularize the UI layer, and various other modules. Complete code can be checked at - https://github.com/Dragoon-Lab/topomath Designed and implemented a Matrix API (in JavaScript) with basic operations like addition, multiplication and inverse (using LU decomposition). This API is being used for solving a system of equations using Newton-Raphson method. Code can be checked at - https://github.com/Dragoon-Lab/topomath-solver .Part of maintenance of Dragoon tutoring system and use Machine Learning techniques to model the student understanding using the data that was collected. Working on assignment of blame problems - changing algorithms for Forward - Backward algorithm in Bayesian Knowledge Tracing (HMM) to incorporate for interaction between knowledge components.

Tempe, Arizona

Worken on design and implemetation Intelligent Tutoring System called Dragoon, which helps student visualize dynamic models (system that change with time http://en.wikipedia.org/wiki/System_dynamics). I have worked on desiging and developing :1. methods to capture student actions, when he is interacting with Dragoon.2. methods to harness these values and create a dashboard for teacher to help them in classroom orchestration.3. assessment of student work using Bayesian Knowledge Tracing. This will be part of my Masters thesis.In parallel I have helped the team in developing the system, and have handled a lot of functionalities in the project. The project can be checked at : http://dragoon.asu.edu/The publications for the project can be checked at - https://dragoon.asu.edu/?q=publicationsGithub repository - https://github.com/Dragoon-Lab/LaitsV3

Gurgaon Sez

I have been part of seven live sites.1. shopadidas.com2. bareescentuals.com & bareescentuals.co.uk3. m.bareescentuals.com4. proactiv.com5. wenhaircare.com6. proactivplus.com7. moet-hennessy-collection.com.auMy responsibilities have grown many folds with every project, and have received many personal as well as team level accolades. During my course of work I have developed as well as designed various projects. Few highlights, In Proactiv I had single-handedly worked on central business module of Kit. In moet-hennesey I was part of the core designing team which had worked on the initial coding structures and data flow in the project.The complete details of my Professional Experience can be checked at : https://sites.google.com/site/sachingrover211/professional

Doctor Of Philosophy (Phd), Automated Planning, Artificial Intelligence, 4.00/4.00

Master Of Science (Ms), Computer Science, 3.71/4

Bachelor'S In Technology, Computer Science

Education record