cv

A short research CV. The full PDF is linked above.

General Information

Full Name Trilok Padhi
Location Atlanta, GA, USA
Email tpadhi1@student.gsu.edu
Research Interests Multimodal agents, large vision-language models, knowledge-enhanced and interpretable AI, AI safety, reinforcement learning for reasoning and planning

Education

  • 2022–now
    PhD in Computer Science
    Georgia State University, Atlanta, GA
    • Advisors and mentors: Prof. Ugur Kursuncu, Prof. Yi Ding, Prof. Valerie Shalin, Dr. Yaman Kumar Singla.
    • Research area: design of multimodal knowledge-enhanced neuro-symbolic models and agents.
    • Coursework: Large Language Models, Advanced Deep Learning, Advanced Machine Learning, Advanced Human-Computer Interaction.
  • 2016–2020
    Bachelor of Technology in Electronics and Communication Engineering
    Veer Surendra Sai University of Technology, Burla, India

Research Experience

  • Summers 2025 & 2026
    Applied Scientist Intern
    Siemens GenAI R&D, Seattle, WA
    • Mentors: Dr. Robert Zhu, Dr. Ivo Santos. Data & AI Research (DAI) team.
    • 2025
      • Development and evaluation of large-scale multimodal industrial foundation models, focusing on robust pretraining strategies and post-training techniques, aligned with Siemens' Industrial Foundation Models initiative.
    • 2026
      • Design and implementation of a novel self- and meta-harness pipeline for synthetic data generation and evaluation for Physical AI foundation models, enabling high-quality synthetic datasets for training and evaluating AI models in industrial applications.
  • Summer 2024
    Research Intern
    SRI International, Menlo Park, CA
    • Mentors: Dr. Susmit Jha, Dr. Anirban Roy, Dr. Ramneet Kaur, Dr. Manoj Acharya. Neuro-Symbolic Computing and Intelligence (NuSCI) research group.
    • Proposed a new state-of-the-art methodology for uncertainty quantification in large vision-language models using visual grounding, published as Seeing is Believing (arXiv:2505.03788).
  • Jan–Aug 2020
    Research Intern
    Bosch Research, Bengaluru, India
    • Mentor: Dr. Sahana Prabhu.
  • 2018–2019
    Research Intern
    NVIDIA Research, Bengaluru, India
    • Mentors: Dr. Christoph Angerer, Mr. Jakob Progsch, Mr. Bharatkumar Sharma, Mr. Jigar Halani.

Industry Experience

  • 2021–2022
    Data Scientist
    Rakuten Inc., Bengaluru, India
    • Supervisors: Mr. Anirban Nandi (Vice President), Mr. Gaurav Ray (Associate Director).
    • Led projects in the Data Science Consulting department on impact estimation, funnel analysis, and customer acquisition, working with a team of 24 across seven countries.
    • Developed an LSTM-Attention + DistilBERT churn prediction model with interpretable attention insights, improving retention strategy with an estimated revenue impact of ¥10M annually.
  • 2020–2021
    Deep Learning Researcher
    Wowexp Technologies, Bengaluru, India

Honors and Awards

  • 2025
    • ACL Student Travel Award
  • 2024
    • IEEE BigData Student Travel Award
  • 2018
    • National Champion (1st of 200,000), Smart India Hackathon, Jaipur, India
  • 2016–2020
    • Academic Scholarship, Veer Surendra Sai University of Technology

Technical Skills

  • Programming and Deep Learning
    • Python (PyTorch, NumPy, Pandas, scikit-learn)
  • Multimodal and Vision-Language Models
    • Hugging Face Transformers, OpenCV
    • CLIP, LLaVA, LLaMA v1/v2/v3, Mistral, GPT-OSS
    • LLM finetuning, pretraining, and post-training (RLHF, PEFT, DPO, alignment)
  • Knowledge Graphs and Graph Neural Networks
    • PyTorch Geometric, NetworkX, DGL
  • High-Performance Computing
    • MPI, CUDA, Slurm

Academic Interests

  • Multimodal agents and large vision-language models
    • Interpretability and early failure detection in long-horizon agent trajectories.
    • Calibration and uncertainty quantification grounded in visual evidence.
  • Knowledge-enhanced learning
    • Commonsense and structured knowledge infusion into compact multimodal models.
    • Neuro-symbolic approaches to multimodal reasoning.
  • AI safety and social computing
    • Red-teaming vision-language models and agents under multi-turn adversarial pressure.
    • Evaluating LLMs in sensitive domains such as mental health support.