Rishav Ray

AI Researcher

Bengaluru, Karnataka, India4 yrs 6 mos experience

Most Likely To SwitchAI ML Practitioner

Key Highlights

Led initiatives to improve personalized search ranking.
Designed a multilingual search engine using Generative AI.
Architected a real-time recommendation system improving engagement by 93%.

Stackforce AI infers this person is a Data Science expert in E-commerce with a focus on recommendation systems and NLP.

Contact

Skills

Core Skills

Recommender SystemsNatural Language Processing (nlp)Machine LearningDeep Learning

Other Skills

Python (Programming Language)PyTorchTensorFlowDatabricksPySparkLarge Language Models (LLM)Neural Ranking ModelsReal-time Vector SearchSemantic SearchQuery CorrectionIntent PredictionAmazon Web Services (AWS)Apache AirflowPersonalizationGenerative AI

About

Experienced Data Scientist with 5+ years in designing and deploying large-scale search and recommendation systems. Skilled in leveraging NLP and computer vision to enhance system intelligence, and proficient in implementing advanced retrieval and ranking strategies that drive user engagement and satisfaction.

Experience

4 yrs 6 mos

Total Experience

1 yr 6 mos

Average Tenure

2 yrs 2 mos

Current Experience

Zepto

2 roles

Senior Data Scientist

Promoted

Apr 2025 – Present · 1 yr 1 mo · On-site

As part of the Search & Recommendations team, I evolved into leading both the Search Relevance and Ranking charters. I lead initiatives focused on improving personalized search ranking, query relevance, and user engagement through deep learning based models, real-time vector search systems and continuous online experimentation. My work spans end-to-end model development, optimization and scalable production deployment, resulting in significant improvements in search quality, CTR, and overall user satisfaction.
Search Ranking:
Hyper personalized search ranking powered by deep representation learning and neural ranking models.
Cohort personalized search ranking powered by custom deep learning based propensity model (Mixture of Experts).
Cold start ranking using Multi Armed Bandits.
Search Relevance:
Real-time semantic search
Two-tower transformer based embedding model
Query correction
Intent prediction
Query segmentation
Attribute understanding
Recommendations:
Cross-sell recommendations including Query-Query and Query-Item recos using MBA, Prod2Vec and GNNs.

Python (Programming Language)Recommender SystemsPyTorchTensorFlowDatabricksPySpark+2

Data Scientist 2

Feb 2024 – Mar 2025 · 1 yr 1 mo · On-site

Nykaa

3 roles

ML Scientist 2

Promoted

Oct 2023 – Feb 2024 · 4 mos

● Designed and implemented the entire personalization stack spanning all Nykaa verticals, launching over 10 new widgets, leveraging Machine Learning and Deep Learning models which significantly increased HP CTR.
● Developed a multilingual and multimodal search engine using Generative AI to upgrade the search experience.
● Crafted the implementation of several deep retrieval and ranking strategies for search and recommendation systems.

PyTorchPySparkAmazon Web Services (AWS)Natural Language Processing (NLP)Apache AirflowMachine Learning+2

ML Scientist 1

Apr 2023 – Sep 2023 · 5 mos

Data Scientist 1

Apr 2022 – Mar 2023 · 11 mos

Trell

2 roles

Data Science Engineer 1

Jul 2021 – Mar 2022 · 8 mos

● Architected and developed a two-stage realtime recommendation system using CLIP, Milvus and ANNOY and deployed it using FastAPI which improved user engagement by 93%.
● Developed DeepFM and vectorsearch based candidate set generators and realtime rankers with Learning To Rank using Redis as a feature store to improve recommendations.
● Formulated an e-commerce product recommendation system using collaborative filtering and deployed it in Kubernetes.
● Built a views prediction system using tag vectors and content metadata.

Data Science Intern

Jan 2021 – Jun 2021 · 5 mos

● Innovated an end-to-end solution for toxic text detection for multiple Indian languages using XLM-Roberta, integrated Kafka, containerized it and deployed it using Kubernetes.
● Created a system to detect trending content using Z-score and setup a cron to update the data regularly to BigQuery.

Sentisum

NLP intern

Oct 2020 – Dec 2020 · 2 mos

● Developed a topic extraction system using transformer-based sentence embeddings, UMAP and HDBSCAN.
● Designed an efficient annotation workflow using Prodigy.

Okcredit

Data Science Intern

Apr 2020 – Sep 2020 · 5 mos · Bengaluru, Karnataka, India

● Created a User Persona system with Graph Networks. Node embeddings were generated using node2vec and GraphSage, then HNSW was used for lookup to predict similar users. Visualization was then done with tSNE.
● Upgraded the category prediction system using transformer models like Roberta and XLNet.
● Designed a machine translation quality estimation system using CNNs, LSTM with attention and improved its performance using transformer architectures like Multilingual Bert, XLM-Roberta and LaBSE.
● Built an OCR system for receipts using Google’s Vision API, then did entity extraction using the Language API.

Indian institute of information technology, allahabad

Research Intern

May 2019 – Jul 2019 · 2 mos

● Developed an intelligent conversational agent with the help of the Wizard of Wikipedia dataset, developed by Facebook AI research using LSTMs with attention mechanism in PyTorch.