I build and optimize large-scale inference systems for foundation models within Amazon's Special Projects incubator (Grand Challenge). My work focuses on the intersection of Generative AI and Biology, specifically engineering systems that handle the computational complexity of multi-billion parameter models in production environments.
My recent research focuses on cross-model translation and prompt optimization—developing methods to reduce inference latency and operational costs while maintaining model reliability. I presented this work at the Amazon Machine Learning Conference (2025).
Research Background
My approach to scalable systems is rooted in algorithmic research across international institutions:
- Universitat Politècnica de Catalunya (UPC), Barcelona: Bachelor's thesis on graph and matrix algorithms for high-dimensional data visualization, supervised by Prof. Ricard Gavaldà and Prof. Marta Arias Vicente. [Thesis]
- University of Florida: Master's in Computer Science. Research on k-tuple harmonious coloring (NP-complete) with Prof. Alper Ungor. Graduate TA for Analysis of Algorithms.
- Tokyo City University: Exchange research under Prof. Takamichi Hirata on algorithmic efficiency within hardware constraints.
- Institute of Mathematical Sciences (IMSc), Chennai: Visiting research student (Top 20 selection, India). Exposure to computational modeling and formal methods.
- IIIT Kancheepuram: Research intern in graph algorithms.
- SASTRA University, India: B.Tech in Computer Science and Engineering.
Scaling Thoughts
This blog documents the engineering decisions behind production machine learning—from low-level inference kernels to the architectural patterns of foundation models.
Opinions expressed here are my own and do not represent my employer.