// ml/ai · data science · data engineering · software engineering

Vanshika Namdev

MS Statistics @ UIUC. I love diving into why a model actually learns, not just what it predicts — and I've got the engineering chops to build it myself.

View Résumé Email LinkedIn GitHub

01

About

I'm drawn to ML and data work because I love the moment a hidden pattern surfaces — that "wait, I didn't see that coming" feeling. I'm just as interested in the math underneath it: the theory of how we optimize, how models actually learn, and how to make them better. I like working on ideas that haven't quite been tried before, and I love that ML and data touch nearly every industry — there's always a new kind of problem to dig into.

Across roles at previous companies, I keep coming back to the same thread: data sits at the core of everything, and how well you understand and prepare it decides what you can actually do with it. I care most about the moments where the data pushes back against intuition and changes a real decision — and about understanding which model, with which hyperparameters, actually fits a given problem, since every round of experimenting teaches you something new.

Outside of work, you'll usually find me dancing or traveling. I love learning about new cultures, trying new food, and meeting people along the way.

02

Skills

Languages

Python SQL Java R

ML / AI

PyTorch TensorFlow / Keras scikit-learn NLP Computer Vision (OpenCV) GenAI / LLM tooling

Data Engineering & Cloud

Airflow PostgreSQL AWS FastAPI Flask Pandas / NumPy

Tools / Other

Power BI DVC Android Studio Git

03

Featured Projects

Farmer's Friend — RAG-based Livestock Triage App

Code

Problem

Veterinary queries over domain-specific livestock corpora needed fast, accurate resolution — a good fit for retrieval-augmented generation, but one that demanded careful pipeline design to be usable in real time.

Approach

Built an end-to-end RAG pipeline in Python using Claude Sonnet, LangChain, and ChromaDB, with Pandas-based preprocessing and optimized chunking for real-time veterinary query resolution over domain-specific livestock corpora.

Impact

Demoed live at the UIUC hackathon to judges from agricultural and ML industry — won first place.

Gen AI / LLMs RAG LangChain ChromaDB Python

LLM-Powered Audit Evaluation Agent

Internal — Mercedes-Benz R&D India

Problem

Compliance teams had to manually process 200+ rulebooks to generate reports, flag high-risk items, and resolve queries.

Approach

Built an LLM-powered AI agent using GPT-4 and a transformer-based RAG architecture to automate report generation, high-risk flagging, and real-time query resolution across 200+ compliance rulebooks — taken from POC to full production.

Impact

60% time less spent 60% time spent on compliance workflows. Gold Award, Mercedes-Benz R&D India

GenAI / LLMs RAG Python Streamlit

SCANR — NEET Risk Prediction

Code

Problem

Buckinghamshire Council needed to identify students at high risk of becoming NEET (Not in Education, Employment or Training) early enough to intervene.

Approach

As 1 of 15 fellows selected globally for DSSGx UK, built a predictive risk model with a version-controlled ML pipeline (DVC) and a Power BI dashboard for council staff.

Impact

Presented at Datafest (The Shard, London) and the LDDG conference. Runner-up, National Data Science & AI Showcase (The Alan Turing Institute).

Python Power BI Pandas / NumPy DVC Machine Learning

// other projects

Sahayak

Android app predicting depression risk by asking users real-life questions.

Android Studio · TensorFlow / TFLite · Keras · Java

Sign Language Detection

Real-time recognition of sign language gestures from video input.

OpenCV · TensorFlow · NumPy · Python

Selp

Flask app predicting mental disorders such as PTSD and anxiety from user input.

Flask · Python · Pandas · scikit-learn

BERT Q/A Flask App

Contextual question-answering over a set of PDF files — exact or most-similar answers.

Flask · CDQA · PyPDF2 · Python

Chatbot

Conversational chatbot built on an intent-recognition dataset from Kaggle.

PyTorch · NLP · NumPy · Python

04

Experience

  1. Fall 2026

    Graduate Teaching Assistant

    University of Illinois Urbana-Champaign — ATMS 517, ATMS 597
    • Grade Python coding assignments across two courses; provide additional instructional support as needed
  2. Jun 2026 — Aug 2026

    Data Science Intern

    SSA Marine
    • Analyzed 4M+ row container dwell-time datasets in Snowflake/Snowpark to inform pricing strategy, contributing to an estimated $5–7M revenue increase
    • Built an AI-powered Market & Acquisition research tool, improving research efficiency by 70%
    • Applied statistical methods (hypothesis testing, confidence intervals, etc) to translate large-scale data into strategic business recommendations
  3. Spring 2026

    Graduate Teaching Assistant

    University of Illinois Urbana-Champaign — ATMS 517, ATMS 526
    • Graded Python-based assignments and updated course materials; contributed additional support work including video captioning
  4. Fall 2025

    Graduate Teaching Assistant

    University of Illinois Urbana-Champaign — ATMS 523
    • Graded assignments and led office hours for a 40-student course; supported students with Python-based coding assignments, debugging, and GitHub-related questions
  5. Aug 2023 — Jul 2025

    Data Scientist

    Mercedes-Benz Research & Development India
    • Built an LLM-powered compliance agent using GPT-4 and a transformer-based RAG architecture to process 200+ compliance rulebooks, automating report generation and high-risk flagging — reducing manual review time by 60%
    • Built an NLP-based tool to automate Engineering Change workflows, improving efficiency by 70% and ensuring structured, clean input data
    • Led development of ML-based classification for part types across carlines, automating 90% of manual effort and improving data integrity
  6. Feb 2023 — Jul 2023

    Data Engineer Intern

    Nunam
    • Designed PostgreSQL & Airflow pipelines at an early-stage climate startup, automating daily ingestion of 10,000+ energy data points and reducing manual processing by 80%
    • Improved ETL data quality by 60% through pandas/NumPy validation scripts for battery analytics models, with database administration in pgAdmin
    • Deployed FastAPI microservices on AWS — integrating Kafka and Postman — enabling real-time data access for 5+ downstream analytics teams
  7. Jun 2022 — Aug 2022

    DSSGx UK Fellow

    University of Warwick, Coventry, UK
    • Selected as 1 of 15 fellows globally to work with Buckinghamshire Council on a data-driven social impact project
    • Developed a predictive model to identify high-risk NEET students, aiming to inform targeted interventions
    • Presented project outcomes at Datafest (The Shard, London) and the LDDG conference
  8. Nov 2020 — Jan 2021

    Graphic Designer

    ClearExam

    Produced design content in Photoshop, responsible for daily design and graphic elements.

  9. Apr 2020 — May 2020

    Python Programmer

    CodeSpeedy Technology Pvt Ltd

    Wrote technical tutorials on Python, OpenCV, and Machine Learning — read them here.

05

Achievements

First Prize — Precision and Digital Agriculture Hackathon

Winner for Farmer's Friend — a RAG-based livestock triage app.

Gold Award — Mercedes-Benz R&D India

Recognized for innovative contributions in business intelligence and GenAI applications.

Winner, Baker Hughes Challenge — TAMU Datathon

Visualized KPIs across 200+ engines, 30+ plants, and 10+ customers using Python and Power BI.

First Prize — Hack-Star Bengaluru

MBRDI × Deccan Herald hackathon. ₹2,00,000 prize plus a pre-placement offer from MBRDI.

Runner-up — National Data Science & AI Showcase

The Alan Turing Institute, hosted by University of Warwick — poster on NEET-risk prediction.

Best ML Hack — Grizzhacks 5

Oakland University hackathon.

Microsoft Learn Student Ambassador

Connected and learned with students globally; hosted and attended technical events.

Core Lead (ML) — Innogeeks

Mentored in the Machine Learning domain at KIET's technical club.

06

Research

Token-Efficient Memory for Agentic Coding

Ongoing — with a CS faculty advisor

Focus

Reducing token usage in multi-turn agentic coding workloads while preserving precise code context

Approach

Designing and evaluating a diff-based structured memory system for coding agents

Status

~2 months in — experiment design & environment setup, running Qwen3-27B on DeltaAI GH200 to compare diff-based vs. full-code context on SWE-bench tasks

AI-Driven Logistics Impact Screening for Engineering Changes in the Automotive Supply Chain

Read Paper

Venue

SAE Technical Paper Series — Paper No. 2025-01-5081

Published

December 2025

Co-authors

Tejas Surampudi, Vishwas Yadav, Tejaswee Anandan

Result

AI-driven optimization achieving 83–85% accuracy

ML Models for Housing Price & Rental Estimation

UC Davis · Jun–Sep 2021

Advisor

Prof. Setareh Rafatirad

Approach

Applied Random Forest and CNN models to estimate property prices and rental values from real-world housing datasets

Contribution

Data cleaning, trend analysis, and visual reporting; also debugged and optimized model integration in the team's Android app

07

Education

MS Statistics

University of Illinois Urbana-Champaign
2025 — 2027

B.Tech, Computer Science

KIET Group of Institutions, Ghaziabad
2019 — 2023 9.16 CGPA Dr. A.P.J. Abdul Kalam Technical University Branch Topper, 2019 — 2020

08

Let's talk

Open to ML/AI, data science, and software engineering roles — and always up for a chat.

vanshikanamdev9@gmail.com LinkedIn GitHub Résumé