AI Research Intern - DIL

AboutGalaxEye

GalaxEyeis an Indian space tech startup dedicated to pioneering multi-sensor imaging satellites, founded by IIT Madras students and alumni and incubated at IIT Madras. The company is developing the Drishti Mission - the world's first satellite with multi-sensor capabilities that deliver all-time, all-weather imagery from space, even during nighttime and in challenging conditions.GalaxEye'stechnology enables advanced geospatial analyses for governments, defense, and industries worldwide.

Our AI team buildscutting-edgedeep learning systems that extract intelligence from high-resolution EO and SAR satellite data. This internship offers an opportunity to work on proprietary datasets and contribute to novel research problems at the intersection of computer vision, multimodal AI, and remote sensing.

What You'll Work On

You'llwork on challenging research problems in satellite image understanding and multimodal geospatial AI. Depending on your background and interests, you may contribute to one or more of the following areas:

• SAR-to-EO image translation using generative models such as diffusion models and GANs.

• Vision-Language Models (VLMs) for understanding, retrieval, reasoning, and interaction with EO and SAR imagery.

• Foundation models for remote sensing, including representation learning, multimodal fusion, segmentation, detection, and large-scale pretraining on satellite imagery.

This is a research-focused role whereyou'llbe expected to read deeply, experiment rigorously, and contribute novel ideas rather than simply implement existing solutions.

What You'll Do

• Conduct structured literature surveys onstate-of-the-artdeep learning, multimodal learning, and remote sensing research.

• Reproduce and benchmark recent research papers on EO and SAR datasets.
• Train, fine-tune, and evaluate deep learning models on proprietary high-resolution satellite imagery.
• Design and execute rigorous experiments with proper tracking, ablations, and analysis.

• Develop and adapt architectures including CNNs, Vision Transformers, diffusion models, GANs, and multimodal foundation models.

• Analyze model performance,identifyfailure modes, and propose improvements.

• Collaborate closely with researchers and engineers, present findings, and contribute to technical reports and research publications.

Requirements

• Strong Python andPyTorch.
• Solid foundations in deep learning and modern computer vision.
• Experience with one or more of the following:

• Transformers and Vision Transformers.

• Generative models (Diffusion Models, GANs, VAEs).

• Vision-Language Models or multimodal learning.

• Familiarity with Linux, Git, and GPU-based training workflows.

• Strong experimental and research mindset.

Good to Have

• Experience with EO, SAR, or geospatial datasets.

• Familiarity with Sentinel-1, Sentinel-2, Landsat,SpaceNet,BigEarthNet, SEN12MS, or related datasets.

• Experience with CLIP, BLIP,LLaVA,SigLIP, Qwen-VL, or similar multimodal models.

• Exposure to raster geospatial data tools such as GDAL orrasterio.

• Experience withexperiment tracking tools such asMLflowor Weights & Biases.

• Publications, research projects, or open-source contributions in computer vision, generative AI, multimodal AI, or remote sensing.

Benefits

• Access to proprietary high-resolution SAR and EO satellite datasets that are unavailable publicly.

• Mentorship from researchers and engineers from top institutions.

• Exposure to the full AI research lifecycle-from problem formulation and data preparation to model development, evaluation, and reporting.

• Opportunity to contribute tocutting-edgeresearch in multimodal geospatial AI.

• Strong performers may have opportunities to co-author publications and be considered for full-time research roles atGalaxEye.