---
title: "Prathamesh Dongarkar - Data Scientist in Bengaluru | Tal"
description: "Meet Prathamesh Dongarkar, Data Scientist @ Meesho | Ex-Amazon Applied Scientist Intern | Applied ML, Personalization & Causal ML | LLMs & GenAI | IIT (BHU) based in Bengaluru. AI stack includes claude code, claude, and codex. Explore public work links including codeforces."
image: "https://tal.club/meet/prathamesh-dongarkar/share-image"
---

# Prathamesh Dongarkar on Tal

## Headline

Data Scientist @ Meesho \| Ex\-Amazon Applied Scientist Intern \| Applied ML, Personalization & Causal ML \| LLMs & GenAI \| IIT \(BHU\)

## About

I am a Data Scientist at Meesho and an IIT \(BHU\) graduate, working at the intersection of Applied ML, personalization, causal ML, and LLMs\. At Meesho, I build production ML systems for User Growth, including contextual bandits, uplift modeling, personalized discounting, and real\-time model serving\. My work has contributed to a 9% ODNR uplift across the discounting and personalization charter\. Previously, as an Applied Scientist Intern at Amazon, I worked on LLM\-based Sponsored Ads ML, including 7B\+ parameter model fine\-tuning, Mixture\-of\-Experts architectures, PEFT, distributed training, and large\-scale inference\. My work included a 137 bps recall improvement and ML systems supporting 1B\+ daily Query\-Ad inferences\. Core areas: Applied ML · LLMs/GenAI · Deep Learning · ML Systems · Deep Learning · Personalization & Recommender Systems · Causal ML · Contextual Bandits I’m currently exploring Applied Scientist, ML Engineer, and Data Scientist opportunities focused on high\-impact ML problems from experimentation and modeling through production\.

## Location

Bengaluru

## AI stack

claude code

claude

codex

cursor

hugging face

openai

## Work experience

Data Scientist at Meesho, May 2025 \- Present · 1 yr 5 mos, Bengaluru, Data Scientist 1 — User Growth, Meesho • Drove a 9% ODNR uplift across the discounting & personalization charter by building personalized discounting using constrained optimization and heterogeneous treatment\-effect modeling\. • Improved personalization across 6 PLP/PDP surfaces by replacing static rules with contextual bandits \(Thompson Sampling / LinUCB\), enabling adaptive card selection based on user context and delayed rewards\. • Designed production\-grade online decisioning and inference, serving ML models through NVIDIA Triton and TensorRT while maintaining online updates to bandit posteriors for real\-time personalization\. • Built end\-to\-end ML data infrastructure using Spark, Databricks, and Airflow for feature engineering, reward attribution, and daily offline feature\-store \(OFS\) tables powering production models\. • Established offline experimentation and model validation frameworks using AUUC, Qini coefficient, uplift/distribution analysis, and bandit traffic/posterior diagnostics to assess incremental impact, convergence, and stability before production rollout

Applied Scientist at Amazon, May 2024 \- Aug 2025 · 1 yr 4 mos, Bengaluru, Applied Scientist Intern & Return Intern — International Machine Learning, Amazon • Improved Sponsored Ads relevance by 4\.2%\+ ROC\-AUC by developing and fine\-tuning transformer\-based models including BERT, LLaMA, Mistral, Mixtral, and Falcon for relevance\. • Improved recall by 137 bps by designing MoE\-BERT, a Mixture\-of\-Experts architecture that increased model capacity by 16× while maintaining similar latency, with custom routing, gating, and expert architectures in PyTorch\. • Scaled fine\-tuning of 7B\+ parameter LLMs across 10 marketplaces, leveraging distributed training and parameter\-efficient techniques including LoRA, QLoRA, and Flash Attention\. • Improved model efficiency and serving economics through INT8/4\-bit quantization while maintaining ROC\-AUC within tolerance, and explored test\-time memory modules for improved long\-context retention\. • Evaluated LLM adaptation strategies across 8\+ regional languages, benchmarking Zero\-/One\-/Few\-Shot and Adaptive In\-Context Learning alongside prompt and fine\-tuning approaches\. • Contributed to ML systems serving 1B\+ daily Query\-Ad inferences across Amazon’s e\-commerce sites and applications, balancing model quality, inference efficiency, and production scalability\.

Amazon ML Summer School at Amazon, Sep 2023 \- Aug 2025 · 2 yrs, India, Apprenticeship

Core Member at Club of Programmers, IIT BHU, Sep 2022 \- Jan 2025 · 2 yrs 5 mos, Varanasi, Uttar Pradesh, India, 𝗖𝗼𝗿𝗲 𝗠𝗲𝗺𝗯𝗲𝗿: 𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 𝗚𝗿𝗼𝘂𝗽 ● Part of the Core Team actively responsible for all the AI and ML initiatives in the Institute\.

Marketing Executive at Kashiyatra, IIT \(BHU\) Varanasi, Dec 2021 \- Nov 2022 · 1 yr, Varanasi, Uttar Pradesh, India

Strategic Relations , E\-Cell IIT\(BHU\) Varanasi at E\-Cell IIT BHU, May 2021 \- Jan 2023 · 1 yr 9 mos, Varanasi, Uttar Pradesh, India

## Education

Integrated Dual Degree\(B\.Tech \+ M\.Tech\), Indian Institute of Technology \(Banaras Hindu University\), Varanasi

Integrated Dual Degree \(B\.Tech \+ M\.Tech\), Metallurgical Engineering, Indian Institute Of Technology\(BHU\),Varanasi

## Public links

- codeforces: https://codeforces\.com/profile/dopamIND\_21

[View the original profile](https://tal.club/meet/prathamesh-dongarkar)

```json
{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Person",
      "@id": "https://tal.club/meet/prathamesh-dongarkar#person",
      "name": "Prathamesh Dongarkar",
      "url": "https://tal.club/meet/prathamesh-dongarkar",
      "image": "https://tal.club/meet/prathamesh-dongarkar/profile-photo.jpg",
      "description": "Meet Prathamesh Dongarkar, Data Scientist @ Meesho | Ex-Amazon Applied Scientist Intern | Applied ML, Personalization & Causal ML | LLMs & GenAI | IIT (BHU) based in Bengaluru. AI stack includes claude code, claude, and codex. Explore public work links including codeforces.",
      "jobTitle": "Data Scientist",
      "worksFor": {
        "@type": "Organization",
        "name": "Meesho"
      },
      "hasOccupation": {
        "@type": "Occupation",
        "name": "Data Scientist"
      },
      "homeLocation": {
        "@type": "Place",
        "name": "Bengaluru"
      },
      "alumniOf": [
        {
          "@type": "EducationalOrganization",
          "name": "Indian Institute of Technology (Banaras Hindu University), Varanasi"
        },
        {
          "@type": "EducationalOrganization",
          "name": "Indian Institute Of Technology(BHU),Varanasi"
        }
      ],
      "knowsAbout": [
        "claude code",
        "claude",
        "codex",
        "cursor",
        "hugging face",
        "openai"
      ],
      "sameAs": [
        "https://codeforces.com/profile/dopamIND_21"
      ],
      "mainEntityOfPage": {
        "@id": "https://tal.club/meet/prathamesh-dongarkar#webpage"
      }
    },
    {
      "@type": "ProfilePage",
      "@id": "https://tal.club/meet/prathamesh-dongarkar#webpage",
      "url": "https://tal.club/meet/prathamesh-dongarkar",
      "name": "Prathamesh Dongarkar on Tal",
      "description": "Meet Prathamesh Dongarkar, Data Scientist @ Meesho | Ex-Amazon Applied Scientist Intern | Applied ML, Personalization & Causal ML | LLMs & GenAI | IIT (BHU) based in Bengaluru. AI stack includes claude code, claude, and codex. Explore public work links including codeforces.",
      "dateModified": "2026-09-04T10:14:58.781Z",
      "isPartOf": {
        "@id": "https://tal.club/#website"
      },
      "mainEntity": {
        "@id": "https://tal.club/meet/prathamesh-dongarkar#person"
      },
      "primaryImageOfPage": {
        "@type": "ImageObject",
        "url": "https://tal.club/meet/prathamesh-dongarkar/profile-photo.jpg",
        "contentUrl": "https://tal.club/meet/prathamesh-dongarkar/profile-photo.jpg",
        "caption": "Prathamesh Dongarkar profile photo on Tal"
      },
      "inLanguage": "en-IN"
    }
  ]
}
```