Applied science · Efficient AI
I lead a team of ML scientists working on efficient AI, connecting machine-learning research with systems that can be deployed in practice.
I am an Applied Science Manager at Amazon, where I lead a team of 10+ ML scientists across India, Europe and the US working on model compression. I previously served as adjunct faculty at IIT (ISM) Dhanbad.
Previously, I founded Transmute AI Lab at IIT (ISM) Dhanbad and co-founded Nyun AI, which was acqui-hired by Amazon in 2025. I have also worked at AIQ, the University of Amsterdam’s QUVA Lab, and Shell Technology Centre. My work connects academic research with applications in industry, including healthcare, energy and subsurface modelling.
I earned my PhD in Computational Engineering at TU Delft. Before that, I completed an integrated B.Sc. and M.Sc. in Geophysics at IIT (ISM) Dhanbad in 2013, receiving the Director’s Gold Medal.
May 2025 — Present
Applied Science Manager, Model Compression
Leads a team of ML scientists within Amazon's ML research organisation, building the internal model-compression platform as part of the hardware–software co-design initiative and working with product, platform and hardware teams on adoption.
Grew a team of 10+ ML scientists across India, Europe and the US, with team research published at CVPR 2026 and ICML 2026.
Aug 2023 — Apr 2025
Co-founder & CEO
Founded and led a venture-backed company that made foundation models deployable on constrained hardware, with ownership across strategy, research, hiring, fundraising, product and customers.
Built Zero and the open-source Nyuntam library, raised pre-seed funding, secured global accelerator cohorts, and led the company through its 2025 acqui-hire by Amazon.
Jun 2021 — Apr 2025
Data Science Manager
Owned data-science delivery across five enterprise programmes spanning reservoir portfolio management, corrosion detection and emissions forecasting.
Jan 2021 — May 2025
Founding Director
Founded the research lab within the TEXMiN Foundation and grew it to around 20 researchers, setting its research agenda and five-year vision.
Jan 2019 — Jan 2021
Postdoctoral Researcher
Worked on model efficiency and visual object tracking in the joint Qualcomm–University of Amsterdam research lab, with applied projects for TomTom, UMC Utrecht and VUmc.
Nov 2017 — Jan 2019
Research Scientist
Applied deep learning to seismic processing and unsupervised event detection in passive microseismic data.
My research focuses on making machine-learning models more efficient, particularly through model compression and efficient inference for foundation models. I also work on computer vision and visual tracking, with earlier research in computational optimisation. My applied work spans healthcare, energy and subsurface modelling.
Selected work through 2025 is listed below. See Google Scholar for the complete publication record.
ACL 2024 · Paper
WACV 2024 · Paper
ICLR 2024 · Tiny Papers · Paper
arXiv preprint · 2024 · Paper
Nyuntam is an open-source PyTorch library for pruning, quantisation, distillation and low-rank decomposition. At Nyun AI, my team also developed Zero, a commercial model-optimisation platform. My patent portfolio includes 2 granted, 1 filed and 3 provisional patents.
Alongside my industry work, I have taught and supervised across several institutions, including 3 PhD, 13 M.Sc. and 20 B.Sc. researchers. My teaching and mentorship connect machine-learning foundations with practical research problems.
IIT (ISM) Dhanbad · Winter Semester 2020–21
Teaching assistants: Rishabh Tiwari, Udbhav Bamba
Course syllabus ↗UiT The Arctic University of Norway
Teaching assistant: Rishabh Tiwari
University of Amsterdam · M.Sc.
Course coordinator and assessor
University of Amsterdam · M.Sc.
Role: Teaching assistant
Delft University of Technology · M.Sc.
Role: Teaching assistant · 2015–16 and 2016–17
Delft University of Technology · Entry-level PhD topic course
Role: Teaching assistant
Delft University of Technology · B.Sc.
Role: Teaching assistant · 2013–14 and 2014–15