Clinical Panda
LLM trained to generate evidence-based explanations for diagnoses using synthetic clinical notes. Introduced lightweight perturbation analysis for real-world robustness.
Welcome!
Machine Learning Researcher with a passion for Large Language Models (LLMs). Leveraging 5 years of software development experience, I've spent the past 2 years specializing in developing, fine-tuning, and deploying LLMs for impactful applications, particularly in healthcare.
Results-driven ML Expert with 5+ years of experience, including 2 years focused on LLMs, RAG, and MLOps. Proficient in building, fine-tuning, and deploying scalable AI systems for real-world applications.
Projects
PII Guard
Enhanced data privacy via advanced PII masking. Involved dataset curation and analysis of model performance under privacy constraints.
Multimodal LLM
Used BLIP-2 and FSDP training to build multimodal LLMs tailored to medical datasets. Focused on efficiency and modality integration.
Neuro-inspired sparse CNN
Designed an interpretable explanation method for CNNs by fusing LRP with sparse coding. Outperformed GradCAM variants in visual domains.
Clinical Tiger
Multimodal RAG system for clinical QA using ClinicalBERT, GPT-4-Vision, chain-of-verification and structured retrieval with citation tracking.
Embeddings to Diagnosis
Robustness evaluation of clinical LLMs via geometric perturbations and PCA-based latent analysis. Identified fragility through boundary crossings.
Skills
- Languages: Python, Scala, Typescript, C, C++, Java, Bash, JavaScript, Ruby
- Web Frameworks: Rails, Flask, Node.js, GraphQL, REST API
- Machine Learning: PyTorch, TensorFlow, LLMs, LangChain, RLlib, Transformers, Hugging Face
- Cloud & DevOps: AWS, GCP, Azure, Docker, Kubernetes, KubeRay, Jenkins, Semaphore
- Developer Tools: Git, Linux, SumoLogic, Shell Scripting, IntelliJ, PyCharm, VS Code
Background
Honors & Awards
- Mitacs Business Strategy Internship (2022)
- PEAK Giver Award – G2 (2020)
- France Travel Grant Recipient (2019)
- National Topper in Mathematics – AISSCE (2015)
Publications
- ARMatron — A wearable gesture recognition glove
RAHA 2016: IEEE Robotics & Automation Society - Clinical Panda: A Large Language Model for Diagnostic Explanation (Under Review)
- Embeddings to Diagnosis: Geometry-Aware Evaluation of Clinical LLM Robustness (In Progress)
Contact
You can get in touch with me through either social media or the below details.