👋 About Me

Hello! I’m Shubham Gaur, a Machine Learning Engineer with 5+ years of experience in Natural Language Processing (NLP), Computer Vision, and MLOps. Currently pursuing my Master’s in NLP at UC Santa Cruz, I’m passionate about developing AI solutions that deliver real-world impact.

💼 My journey in AI has taken me through innovative projects at Adani Group and BlackRock.

🔍 My research interests include NLP, multimodal systems, RAG, and knowledge representation.

🤝 I’m always eager to collaborate on AI projects and explore new ideas. Let’s connect and build solutions that matter!

Work Experience

Lead Machine Learning Engineer @ Adani Group (Nov 2021 - Aug 2024)

AI Solutions for Business Optimization

Built AdaniOne Chat, an LLM chatbot with Delta table-backed vector search.
Deployed GPT-3 models for automated hotel description generation across 33K listings, saving $130K annually.
Developed a churn prediction model using Random Forest & XGBoost, reducing customer churn by 48%.
Applied BERT model to process customer feedback, increasing customer lifetime value by 15%.
Created a recommendation system, increasing duty-free engagement by 87% and boosting revenue by 20%.

Scalable Infrastructure and Data Pipelines

Led a 10-member team to build a real-time Customer Data Platform (CDP), managing data for 100M+ customers.
Conducted cohort analysis on Databricks, driving 50+ personalized campaigns and generating $55K in revenue.

Machine Learning Engineer @ BlackRock (Jul 2019 - Oct 2021)

Developed a neural network-based solution for portfolio and index tree generation, scaling to 100 portfolios across clients.
Built and deployed ETL tools for data validation and securities loading, saving time equivalent to 7 FTEs.
Streamlined financial index summaries, reducing manual effort by 2 FTEs and improving reporting efficiency.

Publications

Advancing Web-Based Visual Question Answering with Efficient Text Alignment

Publication

This paper is a contribution to WebQA, a recent benchmark by Microsoft that combines visual and text reasoning for answers. Exploring alternatives to WebQA’s baseline models like vision language pre-training (VLP) models, my aim was to enhance accuracy through (a) a lighter model, (b) detector-free visual encoder, and (c) knowledge distillation methods such as RoBERTa. Comparing VLP and RoBERTa on single and multi-source questions revealed RoBERTa’s superior performance at 54.91% and 30.11%, respectively, compared to VLP’s 51.49% and 28.73%. This research serves as an advancement to Bing Search.

Extraction of Cumulative Blobs from Dynamic Gestures

Publication

In this project, we’re diving into the world of gesture recognition, a game-changer in the way we interact with computers. Imagine controlling your computer without the need for a mouse or keyboard – that’s the magic of gesture recognition. But here’s the catch – it struggles in low-light conditions due to its reliance on cameras. To tackle this, we got creative and introduced a night vision camera. It’s like giving our system a pair of night-vision goggles, making it thrive where others struggle. We’ve set up a Raspberry Pi with OpenCV, teaching it to spot and track dynamic gestures. With the help of a nifty machine learning algorithm, we’re now recognizing patterns and seamlessly controlling the Raspberry Pi’s GPIOs for various tasks. Guess what? We’ve hit an impressive 99.62% accuracy, proving that even in the dark, our gesture recognition system shines bright.

Research Projects

ADMIRE: Advancing Multimodal Idiomatic Understanding in NLP (Code)
SemEval 2025

Fine-tuned Vision Language Models (VLMs), such as CLIP, to rank images based on idiomatic meanings within context sentences, achieving a top-1 accuracy of 77%.

Question Answering using Retrieval Augmented Generation (RAG) (Code)

Developed a retrieval-augmented QA system using the FLAN-T5 language model and FAISS vector database in LangChain.

Education

University of California, Santa Cruz (UCSC), Santa Clara, CA

Master of Science in Natural Language Processing (Sept 2024 - Dec 2025)

SRM Institute of Science and Technology (SRMIST), Chennai, India

Bachelor of Technology in Information Technology (Jul 2015 - May 2019)

Technical Skills

Programming Languages: Python, C, C++, Java, Perl, SQL, HTML, Node.js
Machine Learning & Deep Learning: PyTorch, TensorFlow, Keras, HuggingFace Transformers, JAX, LoRA Fine-Tuning, CNN
Data Analysis & Scientific Computing: pandas, numpy, scipy, scikit-learn, matplotlib, OpenCV
NLP & LLM Tools: LangChain, HuggingFace Transformers, AutoGen
Development & Deployment: Jupyter, Git, Linux, CUDA, Docker, Kubernetes, FastAPI, Rest API, Streamlit
MLOps & DevOps: MLflow, Airflow, Pinecone, MLOps, CI/CD, Data Factory
Big Data & Distributed Systems: Apache Spark, Kafka, Hadoop, Databricks
Databases & Data Warehousing: PostgreSQL, MySQL, MongoDB, SQL Server, Redis, BigQuery, Synapse
Cloud Platforms: Azure, AWS, GCP
Visualization & Reporting: Tableau, Power BI

Contact

EXTRA CURRICULAR ACTIVITIES

Launched “IT on wheels” initiative to boost computer literacy as a Volunteer for Adani Foundation (NGO), teaching underprivileged students in remote areas of Gujarat post 2022 onwards
Awarded with Employee of the Month in June 2022 for creating a resilient and scalable near real-time data platform.
Accoladed BLK Impact Creator Award for achieving a time save of 10 FTE (Full Time Equivalent) in 2021.
Co-led as Joint Vice President, IT Association, 2018, orchestrating social welfare events in Kattankulathur,aiding 800+ victims of Gaja cyclone in Chennai.