CV

CV

Education

• Ph.D in Computer Science, University of Wisconsin-Madison, 2014
  • Thesis: Efficient Learning of Statistical Relational Models
  • Advisor: Prof. Jude Shavlik Co-Advisor: Prof. Sriraam Natarajan
• M.S. in Computer Science, University of Wisconsin-Madison, 2014
• B.Tech. in Computer Science, National Institute Of Technology (NIT), Tiruchirappalli, 2006
  • Gold Medalist

Work experience

• Lead Research Scientist, Allen Institute for AI, Seattle, WA 2023-Present
  • Building next generation agents for complex reasoning and decision making tasks
  • Building agent benchmarks for tool use, executing experiments, discovery
• Senior Research Scientist, Allen Institute for AI, Seattle, WA 2019-2022
  • Built benchmarks for multi-step reasoning and question answering
  • Built neuro-symbolic models and prompting approaches for complex reasoning tasks
• Research Scientist, Allen Institute for AI, Seattle, WA 2014-2018

• Built benchmarks and symbolic approaches for science question answering

• Software Engineer-II/III, Google R&D Center, Bangalore, India 2006-2008
• Launched local.google.co.in – the first local search (YellowPages) for India

Skills

• Languages: Python, Scala, C/C++, Java, Perl, MATLAB, Shell Scripting, Prolog, HTML
• Technology: PyTorch, Transformers (Huggingface), OpenAI API, AllenNLP, DyNet, Microsoft Scope, Condor Grid Computing, Stanford NLP Toolkit, MapReduce, Big Table

Publications

Olmo: Accelerating the science of language models

AppWorld: A Controllable World of Apps and People for Benchmarking Interactive Coding Agents

Husky: A Unified, Open-Source Language Agent for Multi-Step Reasoning

DiscoveryWorld: A Virtual Environment for Developing and Evaluating Automated Scientific Discovery Agents

DiscoveryBench: Towards Data-Driven Discovery with Large Language Models

Adapt: As-needed decomposition and planning with language models

Bias runs deep: Implicit reasoning biases in persona-assigned llms

Synergpt: In-context learning for personalized drug synergy prediction and drug design

Specializing smaller language models towards multi-step reasoning

Interleaving retrieval with chain-of-thought reasoning for knowledge-intensive multi-step questions

Improving language model negotiation with self-play and in-context learning from ai feedback

How far can camels go? exploring the state of instruction tuning on open resources

Chain-of-Thought Hub: A Continuous Effort to Measure Large Language Models' Reasoning Performance

The Tail Wagging the Dog: Dataset Construction Biases of Social Bias Benchmarks

Teaching Broad Reasoning Skills via Decomposition-Guided Contexts

Retrieval data augmentation informed by downstream question answering performance

Prompt waywardness: The curious case of discretized interpretation of continuous prompts

♫ MuSiQue: Multihop Questions via Single-hop Question Composition

Hey AI, Can You Solve Complex Tasks by Talking to Agents?

Decomposed prompting: A modular approach for solving complex tasks

Complexity-based prompting for multi-step reasoning

Beyond the imitation game: Quantifying and extrapolating the capabilities of language models

Better retrieval may not lead to better question answering

Think you have solved direct-answer question answering? try arc-da, the direct-answer ai2 reasoning challenge

Structure learning for relational logistic regression: an ensemble approach

Prompt waywardness: The curious case of discretized interpretation of continuous prompts

Learning to solve complex tasks by talking to agents

Hey AI, Can You Solve Complex Tasks by Talking to Agents?

GooAQ: Open question answering with diverse answer types

Ethical-advice taker: Do language models understand natural language interventions?

Did aristotle use a laptop? a question answering benchmark with implicit reasoning strategies

Unifiedqa: Crossing format boundaries with a single qa system

UNQOVERing Stereotyping Biases via Underspecified Questions

Text modular networks: Learning to decompose tasks in the language of existing models

Temporal reasoning on implicit events from distant supervision

ReadOnce transformers: Reusable representations of text for transformers

Qasc: A dataset for question answering via sentence composition

More bang for your buck: Natural perturbation for robust question answering

Measuring and reducing non-multifact reasoning in multi-hop question answering

Is Multihop QA in DiRe Condition? Measuring and Reducing Disconnected Reasoning

IIRC: A dataset of incomplete information reading comprehension questions

From ‘F’to ‘A’on the NY regents science exams: An overview of the aristo project

A simple yet strong pipeline for hotpotqa

What's missing: A knowledge gap guided approach for multi-hop question answering

Repurposing entailment for multi-hop question answering tasks

On the Possibilities and Limitations of Multi-hop Reasoning Under Linguistic Imperfections

On the capabilities and limitations of reasoning for natural language understanding

Exploiting explicit paths for multi-hop reading comprehension

Think you have solved question answering? try ARC, the AI2 reasoning challenge

Scitail: A textual entailment dataset from science question answering

Relational restricted boltzmann machines: A probabilistic logic learning approach

Question answering as global reasoning over semantic abstractions

Human-in-the-loop learning for probabilistic programming

Exploiting explicit paths for multi-hop reading comprehension

Can a suit of armor conduct electricity? a new dataset for open book question answering

Bridging knowledge gaps in neural entailment via symbolic models

Adventure: Adversarial training for textual entailment with knowledge-guided examples

Markov logic networks for adverse drug event extraction from text

Learning what is essential in questions

Inductive logic programming meets relational databases: Efficient learning of Markov logic networks

Answering Complex Questions Using Open Information Extraction

Scaling lifted probabilistic inference and learning via graph databases

Question answering via integer programming over semi-structured knowledge

Learning with privileged information: Decision-trees and boosting

Learning continuous-time Bayesian networks in relational domains: A non-parametric approach

Inductive logic programming meets relational databases: An application to statistical relational learning

Combining retrieval, statistics, and inference to answer elementary science questions

Statistical relational learning for handwriting recognition

Learning probabilistic logic models with human advice

Knowledge-based probabilistic logic learning

Graph-based approximate counting for relational probabilistic models

Gradient-based boosting for statistical relational learning: the Markov logic network and missing data cases

Extracting adverse drug events from text using human advice

Exploring Markov Logic Networks for Question Answering

Effectively creating weakly labeled training examples via approximate domain knowledge

Anomaly detection in text: The value of domain knowledge

Relational one-class classification: A non-parametric approach

Relational learning helps in three-way classification of Alzheimer patients from structural magnetic resonance images of the brain

Learning from imbalanced data in relational domains: A soft margin approach

Classification from one class of examples for relational domains

Accelerating imitation learning in relational domains via transfer by initialization

Using commonsense knowledge to automatically create (noisy) training examples from text

Learning relational probabilistic models from partially observed data-opening the closed-world assumption

Bootstrapping Knowledge Base Acceleration.

Structure learning with hidden data in relational domains

Learning Relational Structure for Temporal Relation Extraction.

Gradient-based boosting for statistical relational learning: The relational dependency network case

A machine learning pipeline for three-way classification of Alzheimer patients from structural magnetic resonance images of the brain

Learning markov logic networks via functional gradient boosting

Exploiting causal independence in Markov logic networks: Combining undirected and directed models

Clustering Twitter feeds using word cooccurrence

Boosting relational dependency networks

Adaptive Filetype Aware Prefetching

Some new directions in graph-based semi-supervised learning

How Creative is Your Writing?

Natural perturbation for robust question answering