CV

PhD application CV — Dhwanil R. Chauhan. Graduate Researcher at CIVS, Purdue University Northwest. Multimodal AI · Audio-Visual Learning · Multi-Agent Systems · Industrial AI Safety.

Contact Information

Name Dhwanil R. Chauhan
Professional Title Graduate Researcher · Multimodal AI & Agentic Systems
Email chauha56@purdue.edu

Professional Summary

Graduate researcher at the Center for Innovation Through Visualization and Simulation (CIVS), Purdue University Northwest. My research focuses on multimodal AI, audio-visual learning, and multi-agent systems — building intelligent systems that perceive, reason, and act in real-world industrial environments. Published at CVPR Workshop 2026 and AISTech 2025/2026. Two papers in preparation targeting IEEE TPAMI and ACL.

Experience

  • 2024 - present

    Purdue University Northwest, Hammond, IN

    Graduate Research Assistant
    Center for Innovation Through Visualization and Simulation (CIVS)
    Research at the intersection of multimodal AI and industrial safety, in partnership with the Steel Manufacturing Simulation and Visualization Consortium (SMSVC).
    • Built a conversational AI system guiding supervisors through structured incident capture via natural dialogue, replacing fragmented manual safety reporting across incident types spanning crane failures, pinch points, and equipment damage. (AISTech 2025; V2 targeting ACL)
    • Developed a multi-camera spatial reasoning system dynamically redefining safety zone boundaries from live equipment positions, enabling proactive hazard detection before incidents occur. (AISTech 2026)
    • Contributed core architectural components of a feed-forward framework synthesizing geometrically accurate binaural audio from standard video for industrial safety simulation. (CVPR Workshop 2026)
  • 2025 - 2025

    Remote

    AI Research Analyst
    Untapped Ventures
    • Designed three investment-decision LLM agents (Pre-Seed, Seed, Studio) that ingest heterogeneous pitch materials and generate structured invest/pass recommendations against a formalized multi-stage evaluation rubric.
    • Built an automated screening pipeline with multi-source retrieval and LLM-based scoring (1–100 scale); three-tier confidence routing triggered autonomous outreach above 67, analyst review between 40–67, and batched triage below 40.
    • Iterated through prompt engineering cycles to improve structured output reliability and context retention, developing applied intuitions around LLM failure modes in multi-step reasoning pipelines.

Publications

  • 2026
    Visual Geometry Grounded Novel-View Acoustic Synthesis
    CVPR Workshop 2026

    First unified framework for novel-view acoustic synthesis bypassing explicit 3D reconstruction via feed-forward visual geometry grounding. Contribution: designed the VGGT output representation and formulated the Q/K structure of the Geometry-Grounded Acoustic Decoder (GGAD) cross-attention mechanism.

  • 2026
    Development of Trialing Image Detection for a Melt Shop Safety Tool
    AISTech 2026

    Multi-camera perception system for dynamic safety zone reconfiguration in active industrial environments. Contribution: developed the rule-based spatial reasoning engine fusing four camera perspectives to redefine safety boundaries from outermost detected blocker positions.

  • 2025
    Artificial Intelligence-Assisted Accident Investigation: Improving Safety Reporting with LLMs
    AISTech 2025

    Conversational AI system for industrial safety incident management with context-aware dialogue and dynamic action sequencing. Contribution: designed and implemented the conversational AI backend — monolithic Django pipeline handling context retention, dialogue state, and action sequencing.

  • 2026
    VLM Robustness Benchmark Under Simultaneous Multimodal Degradation
    In Preparation — Targeting IEEE TPAMI / IJCV

    Lead researcher. Systematic evaluation of 20 vision-language models under controlled simultaneous visual and linguistic corruption conditions; novel text corruption module and structured evaluation pipeline.

  • 2026
    AI-Assisted Accident Investigation V2: Multi-Agent Architecture Benchmark
    In Preparation — Targeting ACL

    Architectural evolution from monolithic pipeline to modular multi-agent system; on-premise LLM evaluation benchmark for industrial agentic deployment.

Education

  • 2024 - 2026

    Hammond, IN

    Master of Science
    Purdue University Northwest
    Computer Science
    • Thesis track under Prof. Yang Ni
    • Research focus: Multimodal AI, Audio-Visual Learning, Multi-Agent Systems
  • 2020 - 2024

    Gujarat, India

    Bachelor of Technology
    Charotar University of Science and Technology (CHARUSAT)
    Computer Science
    • 9 publications across ML domains including security, medical imaging, and NLP
    • ACM Student Body President
    • Training and Placement Coordinator

Skills

Research Areas: Multimodal AI, Vision-Language Models, Audio-Visual Learning, Multi-Agent Systems, Industrial AI Safety, Agentic Pipelines, Robustness Evaluation
Frameworks & Tools: PyTorch, Hugging Face Transformers, LangChain, LangGraph, OpenCV, VGGT
Languages: Python, SQL, JavaScript, LaTeX
Infrastructure: Multi-GPU Training, REST APIs, Docker, Git

Academic Service

  • 2024 - present

    Springer Nature

    Peer Reviewer
    Neural Computing and Applications
  • 2024 - present

    UBahrain

    Peer Reviewer
    International Journal of Computing and Digital Systems (IJCDS)

Languages

English : Professional working proficiency
Gujarati : Native speaker
Hindi : Fluent