Research Fellow at Halmstad University·2024 – 2026

Research multimodal AI with an advisor from Meta Llama

I'm working on better scientific question generation, and I pre-trained an any-to-any model on 8TB of data.

Talk to me about
AI Engineer at TU Munich·2025

Used AI agents to make UI test generation 10x faster

At TU Munich and SAP my framework handled flaky tests and added multimodal end-to-end testing.

Talk to me about
Bachelor of Technology - BTech in Computer Science/Artificial Intelligence at Bennett University·2021 – 2025

Earned a 9.68/10 GPA in CS and AI

In 2023 I co-authored a paper on detecting objects by fusing thermal and visible images.

Talk to me about
Research paper·2025

Published at ICLR on helping LLMs reason about space

With Vidushee Vats I worked on spatial reasoning so language models can better locate what an image refers to.

Talk to me about
Researcher at Georgia Institute of Technology·2024

Topped GPT-4V at visual grounding on RefCOCO

Under Dr. Vijay Madisetti we built a framework that takes both images and video, not just one modality.

Talk to me about
Research paper·2024

Outperformed GPT-4 Vision on plant diagnosis

We built a chat system that diagnoses plant disease and released the first multimodal plant-pathology data.

Talk to me about
Computer Vision Engineer at Zocket·2024

Hit 94% accuracy recognizing web ads

I trained dual-encoder models for ad recognition and built a segmentation model for fast background removal.

Talk to me about
Karun Sharma

Karun Sharma⁠

About

I'm a research fellow at Halmstad University, working on multimodal LLMs with Meta Llama advisors.

Experiences

  1. Research Fellow at Halmstad UniversityResearch multimodal AI with an advisor from Meta Llama

    I'm working on better scientific question generation, and I pre-trained an any-to-any model on 8TB of data.

  2. AI Engineer at TU MunichUsed AI agents to make UI test generation 10x faster

    At TU Munich and SAP my framework handled flaky tests and added multimodal end-to-end testing.

  3. Bachelor of Technology - BTech in Computer Science/Artificial Intelligence at Bennett UniversityEarned a 9.68/10 GPA in CS and AI

    In 2023 I co-authored a paper on detecting objects by fusing thermal and visible images.

  4. Research paperPublished at ICLR on helping LLMs reason about space

    With Vidushee Vats I worked on spatial reasoning so language models can better locate what an image refers to.

  5. Researcher at Georgia Institute of TechnologyTopped GPT-4V at visual grounding on RefCOCO

    Under Dr. Vijay Madisetti we built a framework that takes both images and video, not just one modality.

  6. Research paperOutperformed GPT-4 Vision on plant diagnosis

    We built a chat system that diagnoses plant disease and released the first multimodal plant-pathology data.

  7. Computer Vision Engineer at ZocketHit 94% accuracy recognizing web ads

    I trained dual-encoder models for ad recognition and built a segmentation model for fast background removal.

Follow Karun for more...

LinkedInWebsite

© 2026 Delphi · Terms · Privacy

Karun Sharma

Karun Sharma⁠

About

I'm a research fellow at Halmstad University, working on multimodal LLMs with Meta Llama advisors.

Experiences

  1. Research Fellow at Halmstad UniversityResearch multimodal AI with an advisor from Meta Llama

    I'm working on better scientific question generation, and I pre-trained an any-to-any model on 8TB of data.

  2. AI Engineer at TU MunichUsed AI agents to make UI test generation 10x faster

    At TU Munich and SAP my framework handled flaky tests and added multimodal end-to-end testing.

  3. Bachelor of Technology - BTech in Computer Science/Artificial Intelligence at Bennett UniversityEarned a 9.68/10 GPA in CS and AI

    In 2023 I co-authored a paper on detecting objects by fusing thermal and visible images.

  4. Research paperPublished at ICLR on helping LLMs reason about space

    With Vidushee Vats I worked on spatial reasoning so language models can better locate what an image refers to.

  5. Researcher at Georgia Institute of TechnologyTopped GPT-4V at visual grounding on RefCOCO

    Under Dr. Vijay Madisetti we built a framework that takes both images and video, not just one modality.

  6. Research paperOutperformed GPT-4 Vision on plant diagnosis

    We built a chat system that diagnoses plant disease and released the first multimodal plant-pathology data.

  7. Computer Vision Engineer at ZocketHit 94% accuracy recognizing web ads

    I trained dual-encoder models for ad recognition and built a segmentation model for fast background removal.

Follow Karun for more...

LinkedInWebsite

© 2026 Delphi · Terms · Privacy

By using this service, you agree to the Terms of Service and Privacy Policy.