-
July 2024 - Present
Research Associate , AI Research Coop
Institute of Experiential AI at Northeastern University
Boston, USA
As an AI Researcher at the Institute for Experiential AI, Northeastern University, I develop biologically-informed Transformers to predict alternative splicing in long pre-mRNA sequences, integrating biological priors and advanced architectures such as Hyena Layers to outperform state-of-the-art models and enable zero and few-shot generalization. I leverage multi-GPU distributed training and advanced optimization techniques, including Flash Attention, for faster convergence. I adapt self-supervised embeddings to ADAR editing site prediction, demonstrating robust transferability and enhanced biological interpretability. I also pioneer Flow Matching and Re-Basin Canonicalization for large-scale neural weight generation, resolving permutation symmetries and scaling models to 11M parameters with 80% faster convergence than diffusion baselines. Additionally, I apply geometric deep learning to refine flow models, optimizing symmetric latent spaces for uncertainty quantification and reliable AI predictions.
-
January - April 2024
Khoury Graduate Teaching Assistant
Khoury College of Computer Sciences, Northeastern University
Boston, USA
I guide students in Foundations of AI through interactive tutorials, clarifying concepts, and fostering their learning journey. By assisting professors with course development, grading, and creating a collaborative environment, I promote teamwork and inclusivity to ensure students excel in AI fundamentals.
January - April 2024
Graduate Teaching Assistant
-
April 2022 - August 2023
Full Stack ML Engineer
Razorpay
Bangalore, India
I led Golang development for the Optimizer - Payments Team, delivering customer-focused solutions to streamline payment operations. I enhanced payment gateway configurations through rule-based prioritization, containerized microservices with Docker, and deployed them on Kubernetes using Helm. I also implemented robust integration tests, ensuring platform reliability. Notably, I introduced a one-click Paytm wallet feature, boosting user engagement, and mentored an intern to develop a charge-back prediction model using an ANN, achieving impactful results.
-
July 2021 - April 2022
Software Development Engineer
Rebel Foods (Formely Faasos)
Bangalore, India
I led back-end development for the In-order team, managing critical microservices for kitchen staff, food orders, and inventory. I designed scalable schemas for storing order and staff data and implemented a key feature enabling validated order cancellations, reducing revenue loss and ensuring system integrity.
July 2021 - April 2022
Software Development Engineer
-
May 2020 - July 2020
AI Research Intern
Indian Institute Of Information Technology
Prayagraj, India
I conducted a comparative study on GANs for text-to-image synthesis using the Caltech bird dataset, devising a novel evaluation method with t-SNE to analyze overlaps between original and generated images. This research provided nuanced insights into GAN performance and advanced methodologies in AI-driven image synthesis.
-
April 2020 - May 2020
Computer Vision Intern
Ocean Energy
Mumbai, India
I developed a system for detecting traffic rule violations from video inputs, identifying vehicles breaching traffic signals, extracting license plate data, and converting it to text. The system integrated with a database for organized information storage, enhancing traffic monitoring and enforcement.
April 2020 - May 2020
Computer Vision Intern
about me
I am Saumya Gupta, an AI researcher with a passion for harnessing cutting-edge technology to solve complex problems. With a Master's degree in Artificial Intelligence from Northeastern University and extensive experience in AI research and development, I specialize in domains such as Natural Language Processing and generative AI.
I am an AI Researcher at Northeastern University’s Institute for Experiential AI, developing biologically-informed Transformer models for RNA splicing prediction. My work combines biological priors with advanced architectures such as Hyena Layers, achieving state-of-the-art performance and enabling generalization across complex genomic tasks.
I also innovate in Flow Matching and Re-Basin Canonicalization for neural weight generation, improving model alignment and achieving faster convergence than state-of-the-art baselines.
Previously at Razorpay and Rebel Foods, I built backend APIs and deployed generative AI systems at scale. My AI journey began with a sign language detection system, inspiring a lifelong passion for human-centered, transformative AI.
Curious and collaborative, I continue to advance LLMs, RAG systems, geometric deep learning, and Generative AI models while mentoring and delivering impactful AI solutions.
