Distinguished Data Scientist · Walmart Global Tech

Arpit Sharma

Building scalable Agentic AI and Conversational AI systems for retail — with Multi-Agent Workflows, the Google ADK, fine-tuned LLMs, and rigorous LLM-as-a-judge evaluation.

Publications Get in touch
Portrait of Arpit Sharma
Conversational AI
01

About

Computer Scientist working at the intersection of language, knowledge, and reasoning.

Who I am

Computer Scientist with 10+ years designing and deploying scalable conversational AI and agentic systems. 6+ granted/published US patents, NAACL 2025 author, and a strategic leader who mentors cross-functional teams and shapes AI strategy for high-impact retail experiences.

Currently Distinguished Data Scientist at Walmart Global Tech in Sunnyvale, CA — leading the Intelligence team behind Walmart's Shopping Assistant and pioneering multi-agent workflows for use cases like event planning and essentials shopping.

Background

PhD, Computer Science — Arizona State University, 2019. Dissertation: "Towards Understanding Natural Language: Semantic Parsing, Commonsense Knowledge Acquisition, Reasoning Framework and Applications" .

M.S., Computer Science (summa cum laude) — Arizona State University, 2014. Thesis: "Solving Winograd Schema Challenge: Using Semantic Parsing, Automatic Knowledge Acquisition and Logical Reasoning" .

B.Tech., Computer Science (with honors) — College of Engineering Roorkee, 2010.

Agentic AI Multi-Agent Systems Google ADK LLM Fine-tuning RAG LLM-as-a-judge Conversational AI RLHF / DPO / GRPO
02

Experience

A decade of building production AI systems — from semantic parsers and BERT-based NLU to LLM-powered agentic workflows.

  1. Distinguished Data Scientist

    Jun 2025 – Present

    Walmart Associates Inc. · Sunnyvale, CA

    • Lead the Intelligence team behind multiple feature launches and MVPs in Walmart's Shopping Assistant.
    • Designed and implemented multi-agent workflows for innovative shopping use cases — event planning, essentials shopping — using frameworks like the Google ADK.
    • Spearheading industry-research collaborations on low-cost, high-quality open-source LLM judges.

  2. Principal Data Scientist

    Oct 2022 – Jun 2025

    Walmart Associates Inc. · Sunnyvale, CA

    • Provided technical leadership to a cross-functional team of 15+ engineers and scientists across multi-phase ML delivery.
    • Designed an agentic evaluation framework for assessing agentic workflows within a retail-focused chatbot.
    • Defined LLM quality & performance metrics and built an LLM-as-a-judge framework to measure them.
    • Led fine-tuning of open-source LLMs for RAG and code-of-conduct compliance using SFT and DPO.
    • Defined a schema for retail AI agents and shipped multiple types — RAG, Instruct, and Planning agents.
    • Led an academic collaboration that produced a paper accepted at NAACL 2025.

  3. Staff Data Scientist

    Dec 2020 – Oct 2022

    Walmart Associates Inc. · Sunnyvale, CA

    • Developed and fine-tuned a BERT-based NLU model for a conversational AI application.
    • Led a clustering-based, self-serve intent prediction model for multi-domain chatbots using the Universal Sentence Encoder.
    • Drove research on domain-specific adaptation of DistilBERT with universal dependency parsing.
    • Designed API contracts for the core intelligence module powering Walmart's conversational AI platform.

  4. Senior Data Scientist

    Sep 2019 – Dec 2020

    Walmart Associates Inc. · Sunnyvale, CA

    • Trained a Bi-LSTM + CRF model for intent classification and NER in Walmart's shopping assistant.
    • Developed a context-aware deep-learning intent model for multi-turn conversations.
    • Built a text normalization pipeline for ill-formatted user inputs, improving downstream model accuracy.

  5. Data Scientist Intern

    May 2018 – Aug 2018

    Walmart Labs · Sunnyvale, CA

    • Developed semantic parsers for customer queries to improve eCommerce efficiency and response relevance.

  6. Application Support Engineer Intern

    May 2013 – Aug 2013

    MathWorks · Natick, MA

    • Technical support for MATLAB toolboxes — parallel processing, image processing, external interfaces, MEX files.

  7. Systems Engineer

    Sep 2010 – Apr 2012

    Infosys · Mysore, India

    • Developed and supported a financial system for insurance agents using Java, PL/SQL, UNIX, COBOL, and Mainframe.
03

Current Research

Pushing the frontier of agentic AI, LLM evaluation, and explainable language understanding.

Multi-Agent Workflows with Google ADK

Designing and deploying multi-agent systems — including RAG, Instruct, and Planning agents — to power innovative shopping experiences like event planning and essentials shopping. Built on agentic frameworks such as the Google Agent Development Kit (ADK).

LLM-as-a-Judge & Agentic Evaluation

Defining quality and performance metrics for LLMs across retail use cases, and building low-cost, high-quality open-source LLM judges to evaluate both single-turn responses and end-to-end agentic workflows.

LLM Alignment & Fine-Tuning

Fine-tuning open-source LLMs for business-specific RAG and code-of-conduct compliance using SFT, DPO, and exploring GRPO, RLHF, and RLAIF for alignment.

Context-aware Intent Prediction

Using inter-utterance context to predict a user's current action in human–chatbot conversations, improving intent classification in Walmart's shopping assistant.

Read the article

Commonsense Reasoning & Semantic Parsing

Knowledge acquisition and logical-reasoning frameworks underpinning explainable NLU — from the Winograd Schema Challenge to today's agentic systems.

04

Technical Skills

A toolbox built across a decade of NLU, deep learning, and modern LLM-based agentic AI.

Agentic AI & Conversational Systems

  • Multi-Agent Systems
  • Google ADK
  • RAG
  • Prompt Engineering
  • Tool Use & Planning Agents
  • Agent Evaluation

LLMs & Alignment

  • LLM Fine-tuning
  • Transformers
  • SFT
  • DPO
  • GRPO
  • RLHF / RLAIF
  • PPO
  • LLM-as-a-Judge

ML & Deep Learning

  • PyTorch
  • TensorFlow
  • BERT / DistilBERT
  • Bi-LSTM + CRF
  • Universal Sentence Encoder
  • Full ML Life-cycle

Core Technology & Architecture

  • Low-latency AI Systems
  • Fault-tolerant Design
  • GCP
  • Docker
  • API Design

Ethical AI & Evaluation

  • Safety & Privacy
  • Compliance
  • Agent Debugging
  • Quality Metrics

Programming

  • Python
  • Java
  • MATLAB
  • HTML
  • MongoDB
  • Oracle SQL
  • Neo4j
05

Featured Open-source

Tools and libraries built during my research.

K-Parser

A general-purpose Knowledge Parser for English — turning natural language into semantic graphs.

View on GitHub

KNET

Events-based conditional commonsense knowledge — automatic extraction and structured representation.

View on GitHub
06

Selected Publications

A curated selection of peer-reviewed work. Full list on Google Scholar

  1. Bottom-Up Synthesis of Knowledge-Grounded Task-Oriented Dialogues with Iteratively Self-Refined Prompts

    Kun Qian, Maximillian Chen, Siyan Li, Arpit Sharma, Zhou Yu

    NAACL — Main Conference · 2025

  2. Retraining DistilBERT for a Voice Shopping Assistant by Using Universal Dependencies

    Pratik Jayarao, Arpit Sharma

    4th Workshop on Reasoning and Learning for Human-Machine Dialogues · 2021

  3. Improving Intent Classification in an E-commerce Voice Assistant by Using Inter-Utterance Context

    Arpit Sharma

    Proceedings of the 3rd Workshop on e-Commerce and NLP · 2020

    PDF BibTeX

  4. Using Answer Set Programming for Commonsense Reasoning in the Winograd Schema Challenge

    Arpit Sharma

    ICLP · 2019

    PDF BibTeX

  5. Combining Knowledge Hunting and Neural Language Models to Solve the Winograd Schema Challenge

    Arpit Sharma, Ashok Prakash, Arindam Mitra, Chitta Baral

    ACL · 2019

    PDF BibTeX

  6. Semantic Representation and Parsing for Voice eCommerce

    Arpit Sharma, Vivek Kaul, Shankara B Subramanya

    R2K Workshop · 2018

  7. Towards Understanding Natural Language: Semantic Parsing, Commonsense Knowledge Acquisition and Applications

    Arpit Sharma

    IJCAI Doctoral Consortium · AAAI Press · 2016

    PDF

  8. Automatic Extraction of Events-Based Conditional Commonsense Knowledge

    Arpit Sharma, Chitta Baral

    AAAI Workshop on Knowledge Extraction from Text · 2016

    PDF

  9. Generating Semantic Graphs from Image Descriptions for Alzheimer's Disease Detection

    Arpit Sharma, Davy Weissenbacher, Chitta Baral, Graciela Gonzalez

    (In)Coherence of Discourse 3 · 2015

    PDF

  10. Towards Addressing the Winograd Schema Challenge — Building and Using a Semantic Parser and a Knowledge Hunting Module

    Arpit Sharma, Nguyen H. Vo, Somak Aditya, Chitta Baral

    IJCAI · AAAI Press · 2015

    PDF

  11. Identifying Various Kinds of Event Mentions in K-Parser Output

    Arpit Sharma, Nguyen H. Vo, Somak Aditya, Chitta Baral

    NAACL Workshop on Events · 2015

    PDF

  12. An Approach to Solve Winograd Schema Challenge Using Automatically Extracted Commonsense Knowledge

    Arpit Sharma, Nguyen Vo, Shruti Gaur, Chitta Baral

    AAAI Spring Symposium, Logical Formulations of Commonsense Reasoning · 2015

    PDF

  13. Solving Winograd Schema Challenge: Using Semantic Parsing, Automatic Knowledge Acquisition and Logical Reasoning

    Arpit Sharma

    Master's Thesis, Arizona State University · 2014

    PDF

07

Patents

Granted and published patents in conversational AI, semantic parsing, and search retrieval.

  • US 11,960,842

    Methods and apparatus for natural language understanding in conversational systems using machine learning processes

    Pratik Sridatt Jayarao, Arpit Sharma, Deepa Mohan · Issued Apr 16, 2024

  • US 11,715,469

    Methods and apparatus for improving search retrieval using inter-utterance context

    Arpit Sharma · Issued Aug 1, 2023

  • US 11,650,980

    Methods and apparatus for retrieving information via an intermediate representation

    Ghodratollah Aalipour Hafshejani, Arpit Sharma · Issued May 16, 2023

  • US 11,568,007

    Method and apparatus for parsing and representation of digital inquiry related natural language

    Arpit Sharma, Vivek Kaul, Shankara Bhargava Subramanya · Issued Jan 31, 2023

  • App. 17/163,087

    Methods and apparatus for refining a search

    Ghodratollah Aalipour Hafshejani, Arpit Sharma, Sahana Radhakrishna, Haoxuan Chen, Tanmay Peter Kuruvilla, Phani Ram Sayapaneni, Kotla Santhos Kumar · Published 2023

  • App. 63/441,531

    System and method for determining a real-time response based on an understanding of the conversational context

    Arpit Sharma · Applied 2023

08

Service & Awards

Peer review for top NLP venues, talks, and academic recognition.

Peer Reviewer

  • NORA'25 — Workshop on Knowledge Graphs & Agentic Systems Interplay, collocated with NeurIPS 2025
  • EMNLP 2024 — Industry Track
  • COLING 2022 — Program Committee
  • ACL Rolling Review — Reviewer since Oct 2021
  • NAACL 2021 — Program Committee
  • ACL 2020 — NLP Applications track reviewer
  • COLING 2020 — Industry Track Program Committee

Selected Talks

  • ECNLP @ ACL 2020 — Improving Intent Classification in an E-commerce Voice Assistant by Using Inter-Utterance Context. July 2020.
  • ICLP 2019, Las Cruces, NM — Using Answer Set Programming for Commonsense Reasoning in the Winograd Schema Challenge. September 2019.

Awards

  • Graduate College Completion Fellowship — ASU, Spring 2019
  • University Graduate Fellowship — ASU, 2016, 2017, 2018
  • Outstanding Master's Student — ASU Computer Science Department, 2014–15
  • B.Tech with Honors — College of Engineering Roorkee, 2010

Affiliations

  • Graduate Research & Teaching Assistant, ASU School of Computing, Informatics & Decision Systems Engineering (2014–2018)
09

Let's connect

Open to research collaborations, speaking, and conversations about Conversational AI.

arpit7123@gmail.com