Open to AI research engineer roles

Parth K. Thaker

AI Research Engineer

Ph.D.-grade optimization and bandit theory, applied to secure LLM systems.

Currently building privacy-safe LLM workflows at Intuitive Surgical. Before that, a Ph.D. at ASU on graph-based bandits and nonconvex optimization, with research at MERL and IISc.

Nonconvex optimizationGraph-based banditsReinforcement learningSecure LLM systems

View research Résumé (PDF)

Parth K. Thaker — AI Research Engineer — Now — AI Research Engineer @ Intuitive Surgical

Published at

NeurIPSECCVICASSPISIT

7 papers · 1 patent · open-source code

2016
IIT Madras
B.Tech + M.Tech
2022
MERL
Research Intern
2024
ASU
Ph.D. — optimization & bandits
now
Intuitive Surgical
AI Research Engineer

Full timeline & CV →

01 — Research

Selected research

All papers →

Featured

NeurIPS 20222022 CodearXiv

Maximizing and Satisficing in Multi-armed Bandits with Graph Information

A pure-exploration bandit method (GRUB and ζ-GRUB) that uses a similarity graph over arms to find the best arm, or one above a threshold, with far fewer samples than standard UCB.

Read the paper

ECCV 20202020 CodearXiv

Differentiable Programming for Hyperspectral Unmixing using a Physics-based Dispersion Model

Makes a physics-based optical dispersion model differentiable and uses it for end-to-end hyperspectral unmixing, jointly estimating material abundances and physical parameters.

Read

ISIT 20202020arXiv

On the Sample Complexity and Optimization Landscape for Quadratic Feasibility Problems

Proves identifiability and sample-complexity bounds for recovering a complex vector from quadratic measurements, and shows the nonconvex landscape lets gradient descent converge globally from arbitrary initialization.

Read

02 — Projects

Selected projects

All projects →

AI Research Engineer · Intuitive Surgical

Secure, privacy-preserving LLM workflows

Privacy-safe assistants and guardrails that let teams use LLMs over sensitive internal material without data leaving the trust boundary.

LLMsRAGPythonCloud

Lead author & maintainer

GRUB: graph-informed pure-exploration bandits

Open-source bandit algorithms that use a similarity graph over arms to find the best (or good-enough) option with far fewer samples than standard UCB.

PythonNumPyMulti-armed banditsGraph signal processing

03 — Writing

Latest writing

All writing →

The real cost of best-arm identification isn't the arms — it's the clusters

When arms come with a similarity graph and rewards vary smoothly over it, best-arm identification's binding cost drops from the number of arms to the number of clusters.

Why You Can Start Gradient Descent Anywhere — Even on an NP-Hard Recovery Problem

With order-n complex Gaussian quadratic measurements, recovering a vector up to phase becomes both identifiable and solvable by gradient descent from any starting point — two faces of one geometric condition.

One Coin Flip That Spans Fast Search and Cheap Search

Two robotic-search strategies optimize opposite costs; randomly interleaving them with one tunable coin lets you slide between "find targets now" and "spend the least energy."