May 2024 – Aug 2024 · 3 mos · Puducherry, India · Hybrid

• Processing millions of protein and chemical sequences data, training language models from scratch.
• Fine-tuning them using techniques like Low-Rank Adaptation (LoRA), instruction tuning, and domain adaptation through continual learning.
• I also explored different quantization levels and optimized training on multi-GPU environments.
• Using LLMs to extract embeddings.
• Developing custom LLMs on top of pre-trained models for protein and drug compound related tasks, learned a lot about optimizing LLMs for specific tasks.

Jul 2023 – Dec 2023 · 5 mos · Bengaluru, Karnataka, India · Hybrid

• Developed robust object detection and tracking solutions using satellite imagery and videos.
• Worked in collection, pre-processing and interpretation of the spatial data using various tools.
• Customizing , Fine-Tuning Computer Vision Models for object detection in satellite imagery.
• Training and Fine-tuning models for multi-object tracking in satellite video sequences.