Sumit Kumar

AI Researcher

Bengaluru, Karnataka, India7 yrs 4 mos experience

AI ML PractitionerAI Enabled

Key Highlights

Expert in multimodal AI and LLMs for innovative solutions.
Proven track record in reducing fraud detection false positives.
Strong background in NLP and deep learning applications.

Stackforce AI infers this person is a Data Scientist specializing in AI/ML solutions for Retail and SaaS industries.

Contact

Skills

Core Skills

Computer VisionNatural Language Processing (nlp)Generative AiDeep LearningData Science

Other Skills

PresentationsGenerative AI ToolsData WranglingCommunicationPredictive ModelingMLOpsPostgreSQLPrompt EngineeringData StructuresAlgorithmsAnomaly DetectionExtract, Transform, Load (ETL)Decision TreesBERT (Language Model)Matplotlib

About

Data Scientist with experience in Natural Language Processing, Computer Vision and Forecasting seeking new challenging opportunities to explore cutting-edge technologies. My experience in these areas, combined with a quick learning ability and strong motivation, enables me to consistently deliver impactful results that drive business growth. I am passionate about leveraging my skills to make meaningful contributions to any project or organization I work with to create State-of-art AI/ML solutions.

Experience

7 yrs 4 mos

Total Experience

2 yrs

Average Tenure

1 yr 2 mos

Current Experience

Walmart

Data Scientist III

Mar 2025 – Present · 1 yr 2 mos · Bengaluru, Karnataka, India · On-site

Productionizing deep learning pipelines integrating RT-DETR, CLIP, and V-JEPA for fine-grained behavioral analysis.
Drove end-to-end design and deployment of multimodal AI systems for real-time retail fraud and theft detection at scale.
Engineered scalable fraud scoring systems combining embedding-based anomaly detection and supervised learning models.
Optimized model precision and business impact, reducing false positives by 28% and unlocking multimillion-dollar savings.

Computer VisionNatural Language Processing (NLP)

Fractal

Data Scientist

Jul 2021 – Apr 2025 · 3 yrs 9 mos · Mumbai, Maharashtra, India · Remote

Designed & executed multilingual LLM experiments, comparing 10+ designs across 18+ metrics. Presented findings & recommendations, informing architectural decisions for 1.5M+ user chatbots.
Developed Streamlit applications showcasing LLM automation for coding (10-100x efficiency gain) & text-to-SQL conversion.
Built Django backend APIs & RAG framework for persona-based chatbots. developed advanced NLP algorithms for contextualizing. 100+ word documents.
Designed & developed a deep learning module for automatic data usage verification ( achieving 70% automation).
Implemented web scraping & data wrangling to curate training data, and deployed a deep learning intent classification model on Azure Databricks.
Increased automation by ~100% through intent classification of legal documents, optimizing code for 5x faster execution.
Mentored interns on NLP & Computer Vision for submodule development and POCs.
Nominated for "Client Champion Award" for outstanding work and contributions.

PresentationsGenerative AI ToolsNatural Language Processing (NLP)Generative AI