Research Scientist · Amazon

Learning to make
better decisions
in a changing world.

I’m Bhargav Ganguly. I work at the intersection of machine learning, sequential decision-making, and optimization—from theoretical guarantees to systems operating under real-world uncertainty.

Explore my research Google Scholar

Purdue PhD ’24 IIT Kanpur ’19 Bellevue, WA

Reinforcement learningDistributed intelligenceGenerative modelsML for operations

01 / Research

Decisions under uncertainty.
Grounded in theory.

How can learning systems adapt when data shifts, communication is limited, and decisions have consequences?

Contours of f(x,y)=(3x²+y²)/2 with a copper gradient-descent path from (2.7,2.4) toward the minimum at (0,0).

A mathematical lens

Small steps.
Better decisions.

Each contour joins points with the same objective value. Each copper dot takes a step against the gradient, approaching the minimum.

f(x, y) = ½(3x² + y²)θt+1 = θt − η∇f(θt)

Illustrative quadratic objective · θ = (x, y), η = 0.14
A view of optimization, not a result from a paper.

01

Learning under change

Federated optimization and concept-drift adaptation for distributed data that doesn’t stand still.

Explore MASTER-FL
02

Sequential decision-making

Multi-agent and quantum reinforcement learning, with regret and convergence guarantees.

Explore quantum RL
03

Efficient distributed learning

Reducing communication in parallel RL and adapting where training and aggregation happen across edge networks.

Explore parallel RL

Research spotlights

Google Scholar
Cooperative edge-assisted federated learning architectureView full-size image

IEEE/ACM ToN 2023 Edge ML · Distributed optimization

Multi-Edge Server-Assisted Dynamic Federated Learning with an Optimized Floating Aggregation Point

Where should distributed learning happen? CE-FL combines device and edge-server training with data offloading and an aggregation point that can move between rounds, supported by convergence analysis and distributed resource optimization.

Bhargav Ganguly*, Seyyedali Hosseinalipour*, Kwang Taik Kim, Christopher G. Brinton, Vaneet Aggarwal, David J. Love, Mung Chiang

Visual overview ↗Read paper Preprint ↗* Equal contribution
Explore five more publications

Research notes / Behind the papers

The ideas behind
the results.

Explanations of the problems, methods, and limits behind my published research.

Explore all four research notes

02 / In practice

From mathematical ideas
to operational decisions.

Research shaped by uncertainty in logistics, distributed learning, and financial systems.

  1. Oct 2024 — Present

    Now

    Amazon

    Research Scientist · Last Mile Routing & Planning

    Bellevue, Washington

    Building ML and optimization systems for delivery capacity planning, with a focus on forecasting uncertainty and understanding driver decisions.

  2. May — Aug 2023
    Jun — Aug 2022

    During PhD

    Amazon SCOT

    Research Scientist Intern · Two summers

    Bellevue, Washington

    Deep reinforcement learning for inventory replenishment, and inference models for vendor behavior.

  3. Jan 2021 — Dec 2024

    Research & education

    Purdue University

    PhD · Industrial Engineering

    West Lafayette, Indiana

    Federated learning under changing data, decentralized reinforcement learning, and optimization. Graduate research and teaching assistant.

  4. Jul 2019 — Dec 2020

    Industry

    JPMorgan Chase & Co.

    Quantitative Developer · Wholesale Credit Core Analytics

    Mumbai, India

    Statistical and ML models to forecast credit losses under economic stress.

  5. May — Jul 2017

    During IIT Kanpur

    Fuzzy Logix

    Data Science Intern

    Bengaluru, India

    Parallel machine learning and cross-validation directly within relational databases.

  6. Jul 2014 — Jun 2019

    Education

    IIT Kanpur

    B.Tech. & M.Tech. · Electrical Engineering

    Kanpur, India

    B.Tech. 2014–2019 · M.Tech. 2017–2019. Research in compressed online multi-class kernel bandit learning.

03 / Foundations

A research mindset.
An engineering foundation.

Industry experience in finance and logistics informs my work on learning algorithms that account for changing data and operational constraints.

Education & academic research

2021 — 2024

PhD · Industrial Engineering

Purdue University

Advisor: Prof. Vaneet Aggarwal

Federated Learning Amidst Dynamic Environments ↗

Research in online federated optimization, decentralized reinforcement learning, and non-convex optimization under uncertainty.

2017 — 2019

M.Tech. · Electrical Engineering

Indian Institute of Technology Kanpur

Advisor: Prof. Ketan Rajawat

Thesis: Compressed Online Multi-Class Kernel Bandit Learning.

2014 — 2019

B.Tech. · Electrical Engineering

Indian Institute of Technology Kanpur

Teaching & research community

8 semesters1,000+ students

Graduate Teaching Assistant at Purdue, teaching and mentoring across stochastic modeling, operations research, quality engineering, and predictive analytics.

Academic service

Program committee and reviewing service includes AAAI, IJCAI, ECAI, ICML, ICLR, and IEEE Transactions on Wireless Communications.

Session Chair, “Coordinated Planning in Complex Transport Networks,” INFORMS Annual Meeting 2025.

Recognition

  • 2024Purdue Graduate School Summer Research Grant
  • 2022Passed CFA Level I examination
  • 2018–19IIT Kanpur Graduate Fellowship
  • 2017IIT Kanpur Academic Excellence Award
  • 2013KVPY National Fellowship

04 / Connect

Let’s talk about
what comes next.

For conversations about AI research roles, scientific collaboration, or the work featured here, connect with me on LinkedIn.

Connect on LinkedIn Google Scholar GitHub