Shubhendu Sharma

Data Scientist

Bengaluru, Karnataka, India8 yrs 11 mos experience

AI EnabledAI ML Practitioner

Key Highlights

7+ years of experience in AI and machine learning.
Expert in building scalable AI systems for real-world applications.
Proven track record in predictive modeling and generative AI.

Stackforce AI infers this person is a Data Scientist specializing in AI and machine learning solutions for enterprise applications.

Contact

Skills

Core Skills

Machine LearningGenerative AiPredictive ModelingData VisualizationData Analysis

Other Skills

PythonFastAPIMongoDBAWSLangGraphLangChainPostgreSQLPandasNumPyGLMsRandom ForestCollaborative FilteringK-Means ClusteringDBSCANQlik-View

About

Sr. Data Scientist | AI & LLM Engineer | 7+ Years in Data Science & Machine Learning I’m a Sr. Data Scientist with over 7 years of experience building end-to-end AI and machine learning solutions across insurance, retail, and enterprise platforms. My work focuses on translating complex data problems into scalable systems that deliver measurable business impact, including pricing optimization, risk prediction, and customer analytics. Recently, my focus has expanded to Generative AI, large language models, and agentic systems, where I design production-grade solutions such as Text2SQL multi-agent frameworks, LLM-powered automation pipelines, and vision-language model based video analytics platforms. I enjoy building AI systems that move beyond experimentation into real-world production, combining strong machine learning foundations with modern AI architectures like RAG pipelines, LangGraph workflows, and multi-agent orchestration. Core Skills: Python, SQL, PySpark, Pandas, NumPy, FastAPI, MongoDB, AWS, Azure Databricks Machine Learning: Predictive Modeling, Pricing Analytics, Recommendation Systems, Anomaly Detection, Time Series, Feature Engineering Generative AI & NLP: LLM Applications, LangChain, LangGraph, RAG Systems, Multi-Agent Architectures, Prompt Engineering, Transformer Models (BERT, GPT) Explainable AI: SHAP, LIME, Model Interpretability Always interested in building impactful AI systems and collaborating on challenging data and AI problems. Let’s connect.

Experience

8 yrs 11 mos

Total Experience

3 yrs 8 mos

Average Tenure

1 yr 7 mos

Current Experience

Nagarro

Senior Data Scientist

Nov 2024 – Present · 1 yr 7 mos

[Hybrid Video Analytics Compliance System]
Architected a real-time SOP compliance monitoring platform using hybrid edge-cloud architecture with VILA-7B VLM for video narration and dual-tier compliance evaluation (Llama-3B on edge, GPT-4o on cloud).
Designed event-driven data pipelines with FastAPI, MongoDB, MQTT, WebSockets; deployed on AWS with site-to-site VPN for on-premise camera integration
[Text2SQL Agent Framework]
Built a production-grade natural language to SQL system using LangGraph's multi-agent architecture. The pipeline orchestrates 6 specialised agents — Entity Resolution, Rule Injection, Query Correction, SQL-RAG Generation, Validation, and Execution — enabling business users to query databases conversationally.
Key innovations: RAG-based schema filtering to reduce context noise, fuzzy entity mapping (e.g., "Europe" → "Western Europe"), and automated SQL validation before execution.
Tech: Python · FastAPI · LangGraph · LangChain · OpenAI · PostgreSQL · AWS ECS

PythonFastAPIMongoDBAWSLangGraphLangChain+3

Chubb

Senior Data Scientist I

Mar 2022 – Nov 2024 · 2 yrs 8 mos · On-site

Asia BTA: Helped the business by creating a model which predicts premium rates using a mix of internal & external variables such that customer is asked at max 2 questions and we can provide a quote for them. Helped the brokers to fasten the process of getting more business for the company. As a result, we were able to generate 20% more quotes per day.
Algorithm Used: GLMs H20.ai
Asia GPA: Created a model that forecasts optimal pricing rates for insured people which was done earlier using manual effort by the actuaries and underwriters. Model uses predictive modelling algorithm for giving out premium rates for pure occupation class segment customers.
Algorithm Used: GLMs H20.ai
Cross Sell: Built a recommendation system which uses class weights for different categories on top of recommendation system and gives out sensible product recommendations for P&C
and Financial LOB. We were able to increase our cross-sell propensity to 40% which was previously 8%.
Algorithms Used: Random Forest, Collaborative Filtering, KNN.
D&O: Responsible for extraction of financial data for Public Companies and implementing the same for Large Private companies and for APAC region. Made the extraction process simpler by using API code instead of using Excel plugins. Also helps in creating peer grouping ( grouping similar companies) using their MarketCap and GICS code.
Tech Stack: Python, Pandas, NumPy.

PythonPandasNumPyGLMsRandom ForestCollaborative Filtering+2

Capgemini

Associate Consultant

Jul 2017 – Mar 2022 · 4 yrs 8 mos · Bengaluru, Karnataka, India

890 by Capgemini: Built an intuitive dashboard for the interpretation of Revenue generated monthly, quarterly, and yearly. Made use of unsupervised algorithms to identify customer segments and based on the customer segments were able to increase the retention rate by 18%.
Algorithm used: K-Means Clustering, DBSCAN.
AIRBUS-ABDAS: Monitoring and building of process chains on BW and over-viewing them through the different landscapes (Development, Testing and Production). Generating reports in Qlik-View to present to clients/Users for them to take valuable decisions. Analyzing data related issues at different layers of data warehousing stage.

PythonK-Means ClusteringDBSCANQlik-ViewData VisualizationData Analysis