Machine Learning Engineer & Researcher

Prarthana Krishnamurthy

I train and ship models that hold up outside the notebook, from diffusion models and fine-tuned LLMs to predictive models for drug discovery.

Most recently trained and deployed molecular property prediction models at Alkermes, owning the lifecycle from feature pipelines and training to evaluation and serving. Before that, trained text diffusion models on 8× H100s at Northeastern and built data systems at Société Générale.

View projects Resume About me →

Highlights

Credentials, recommendations, and milestones

Scroll to explore

Letter of Recommendation

Prof. Dino Konstantopoulos, Teaching Professor, Northeastern University

View Full PDF ↗

          
          Master's Thesis
          ProQuest Publication

Master's Thesis Publication

Non-Autoregressive Text Generation Using Diffusion Models, ProQuest 2025

View on ProQuest ↗

Master of Science in Information Systems

Northeastern University, College of Engineering, December 2025

View Full PDF ↗

Research

Master's Thesis at Northeastern University

Non-Autoregressive Text Generation Using Diffusion Models

Advisor: Prof. Dino Konstantopoulos | Jan 2025 – Dec 2025

Designing continuous-embedding diffusion models for parallel text generation in BERT embedding space, achieving significant improvements in semantic fidelity and inference speed through architectural innovations and advanced sampling techniques.

0.93 BERT Similarity ROUGE-L: 0.742 BLEU-4: 0.267

PyTorch Diffusion Models BERT DPM-Solver++ U-Net

Core Expertise

Where research meets production-grade engineering

Diffusion Models & Generative AI

Expert in designing, training, and optimizing diffusion models for text and image generation.

DDPM DPM-Solver++ U-Net continuous embeddings

Large Language Models & NLP

Deep expertise in training, fine-tuning, and scaling large language models from scratch.

Transformers LoRA/QLoRA RLHF/DPO distributed training

Life Sciences AI

Applying deep learning across medical imaging and molecular modeling, from image synthesis and segmentation to compound property prediction and drug-like molecule generation.

U-Net segmentation PSNR/SSIM GCN/GAT SMILES generation RDKit

MLOps & Cloud Infrastructure

Production-ready ML deployment with containerization, orchestration, and monitoring.

Docker Kubernetes AWS/Azure CI/CD

Let's Build Something Amazing

Open to ML Engineer, Applied Scientist, and Forward Deployed Engineer roles in generative AI, computer vision, and production ML systems

Get In Touch