Hey π β grab a coffee and let’s talk shop.
I’m Nikhil Saini, Chief Engineer on the Language AI team at Samsung Research Institute Bangalore. I work at the intersection of NLP and ML, with a focus on making language models actually useful.
These days I’m thinking about diffusion models for text and agentic systems β two areas I think are underexplored and worth serious attention. If you’re working on either, I’d love to compare notes.
Previously: My background spans Machine Translation, Disfluency Correction, On-device LLMs, Quantization, and Reinforcement Learning β mostly in the text domain. During my post-gradπ¨βπ at IIT Bombay, I worked at the CFILT and CSALT lab with Prof. Pushpak Bhattacharyya and Prof. Preethi Jyothi,tackling problems at the boundary of Speech and NLP.
PC Member: LREC 2022, ICON, LREC-COLING 2024, COLING 2025, LREC 2026.
M.Tech (Research), 2021
IIT Bombay
B.E, 2018
MSU Baroda
Domain:
Domain:
Responsibilities include:
Using denoising auto-encoder and backtranslation, this work beats a baseline model by a margin of 4.21 and 3.11 BLEU points where the baseline uses disfluent and fluent references, respectively, without an access to disfluent-to-fluent parallel corpus.