I’m a Senior Machine Learning Engineer at Google, building on-device AI
features for Android — including Scam Guard (real-time scam detection via an on-device
Gemini model) and New Generation Call Screen. My technical interests centre on LLMs,
Recommender Systems and efficient model training.
Before Google I worked at LinkedIn on AI foundations for recommendation and language models,
mainly exploring techniques for fast and efficient large-batch training. I did my Master’s at NYU
Courant, where I worked with Prof. Brenden Lake
on modelling infant cognitive development using self-supervised deep learning — work published at NeurIPS 2020.
With my better half, Sidvita
Experience & Education
2024 –
Google
Senior Machine Learning Engineer · Mountain View
Building on-device AI features for Android’s Dialer app: Scam Guard (real-time scam detection via on-device Gemini) and New Generation Call Screen.
End-to-end LLM model development: synthetic data generation, labelling, LoRA fine-tuning, and quantization for on-device deployment.
Leading a research reading group on LLMs, fine-tuning, and quantization.
2022 – 2023
LinkedIn
Applied Scientist · Sunnyvale
Developed foundational techniques that could be applied to Personalization and Language Models throughout LinkedIn.
Improved training speed and AUC for Feeds and Ads models via gradient norm clipping, XLA, and mixed-precision training.
Built an Explore-Exploit framework using Thompson Sampling over neural network output variance for recommendation diversity.
2021
Amazon
Applied Scientist · Seattle
Causal Inference to quantify long-term downstream impact of seller actions (sponsored ads, inventory restocking) on revenue.
