Arun Prasath Jayaprakash profile photo 1 on Tal

Arun Prasath Jayaprakash

AI/ML @ 3M | Ex-Ekaya AI, Shell, HTC | MS @ California State University

Bengaluru

5+ years of work experience

3M
Jan 20269 mosPresent

Bengaluru

Ekaya AI
May 20259 mosJan 2026

Santa Monica, California, United States

Architechted multi-agent AI systems using LangGraph and LangSmith to automate database ontology extraction, natural language-to-SQL generation, and intelligent schema analysis across enterprise data warehouses (PostgreSQL, ClickHouse, BigQuery, Snowflake), enabling secure AI access to data. Built a semantic query platform with dual interfaces - MCP servers and FastAPI services leveraging LangGraph workflows for ambiguity detection, query planning, SQL generation, validation, and progressive learning that adapts from user patterns for better sql generation yielding accurate analysis. Developed a high-performance data access layer using Apache Arrow Flight SQL (SDAP) for 10–100x faster analytics for sql based query execution, utilized Redis-based distributed caching (schema, ontology, syntax), ChromaDB vector stores for semantic search, and LLM connection pooling with dynamic model selection based on query complexity enabling faster responses for less complex queries. Engineered ontology and model definition pipelines with automated schema relationship discovery agents that generates ontology based on human and database context to create RDF/OWL files for semantic sql generation based on enterprise data.
Shell
May 20221 yr 4 mosAug 2023

Greater Chennai Area - India

- Enabled finance stakeholders to get transparency and act on bottlenecks for finance invoice workflows by instructing supervised machine learning models (XGBoost, Pytorch models) that forecast bottlenecks with 89% accuracy. - Authored simulation scripts (First In First Out) which give stakeholders a real-time simulation on changes to timeline and invoice workflow patterns, enabling a Power BI dashboard to visualize the impact of removing forecasted bottlenecks. - Derived 38% efficiency in preventing late payments made to vendors using A/B Testing, which was the result of better forecasting from AI models on bottlenecks and explained AI decisions to business with LIME and SHAP tools. - Automated data engineering pipelines (ETL, SQL) for data transformations needed to support AI models using Google Dataproc, with compliance to finance data auditing and region-based data usage policies. - Scaled AI forecasting solution to different stakeholder regions using Google Cloud based Kubernetes, Docker and Terraform scripts, CI/CD with GitHub actions, enabling stakeholders to reduce late payment losses across organization. - Contributed to campaign analysis for shell downstream - lubricants, with interactive dashboards (PowerBI) that enabled stakeholders to get marketing campaign insights and metrics from various data sources. - Enabled faster calculation of campaign metrics (CTR, ROAS, impressions, conversions) for dashboards by approximate 40% with integration of real-time data services managed with Kafka
HTC Global Services
Mar 20193 yrs 3 mosMay 2022

Guindy, Chennai - India

- Trained and productionized an image segmentation model (PyTorch, Deep Learning) to automate the assessment of vehicle damage in auto-insurance claims, reducing processing times and cost by 35% - Collaborated with multiple teams for exposing AI models to end-users via Flask REST API’s, enabling users to get instant access to their insurance quotes, model features and improve automation efficiency for insurance clients via mobile/web. - Enabled faster user responses by moving model to CUDA, improving user mobile application experience on scale. - Deployed the automation solution for client users to consume over the internet with AWS cloud deployment (Docker, Kubernetes), hosting the AI model using sagemaker, monitored with cloudwatch enabling services with no downtime. - Developed a chatbot assistant to give employees under the organization context relevant information on organizational policies, replacing third party services for chatbot systems which reduced 25% cost to company. - Streamlined data pipelines needed for assistants NLP model (BERT - Hugging Face) using Spark distributed services, which enabled larger data handling, better system processes improving reliability when data is stored back to Google cloud storage.

Arun’s AI Stack

claude code
claude code
hugging face
hugging face
kiro
kiro
openai
openai
Microsoft Copilot
Microsoft Copilot
vLLM
vLLM

Education

California State University, Long Beach

Master of Science - MS, Computer Information systems (Data Science & AI)

California State University, Long Beach

20232 yrs2025
SRM IST Chennai

Bachelor of Engineering - BE, Mechanical Engineering

SRM IST Chennai

20134 yrs2017

Join Arun and 10k+ other professionals on

Tal Boss

Tal is the app where bosses meet engineers & builders directly, without middlemen

Download Tal Boss

Arun Prasath Jayaprakash on Tal

Headline

AI/ML @ 3M | Ex-Ekaya AI, Shell, HTC | MS @ California State University

About

Experienced Professional in AI , Machine Learning, LLMs for solving key business problems across industries like finance, marketing, energy and consulting. Proficient in analyzing problems, architechting solutions with AI, LLMs.

Location

Bengaluru

AI stack

claude code

hugging face

kiro

openai

Microsoft Copilot

vLLM

Work experience

Senior AI/ML Engineer at 3M, Jan 2026 - Present · 9 mos, Bengaluru

AI Engineer at Ekaya AI, May 2025 - Jan 2026 · 9 mos, Santa Monica, California, United States, Architechted multi-agent AI systems using LangGraph and LangSmith to automate database ontology extraction, natural language-to-SQL generation, and intelligent schema analysis across enterprise data warehouses (PostgreSQL, ClickHouse, BigQuery, Snowflake), enabling secure AI access to data. Built a semantic query platform with dual interfaces - MCP servers and FastAPI services leveraging LangGraph workflows for ambiguity detection, query planning, SQL generation, validation, and progressive learning that adapts from user patterns for better sql generation yielding accurate analysis. Developed a high-performance data access layer using Apache Arrow Flight SQL (SDAP) for 10–100x faster analytics for sql based query execution, utilized Redis-based distributed caching (schema, ontology, syntax), ChromaDB vector stores for semantic search, and LLM connection pooling with dynamic model selection based on query complexity enabling faster responses for less complex queries. Engineered ontology and model definition pipelines with automated schema relationship discovery agents that generates ontology based on human and database context to create RDF/OWL files for semantic sql generation based on enterprise data.

Data Scientist at Shell, May 2022 - Aug 2023 · 1 yr 4 mos, Greater Chennai Area - India, - Enabled finance stakeholders to get transparency and act on bottlenecks for finance invoice workflows by instructing supervised machine learning models (XGBoost, Pytorch models) that forecast bottlenecks with 89% accuracy. - Authored simulation scripts (First In First Out) which give stakeholders a real-time simulation on changes to timeline and invoice workflow patterns, enabling a Power BI dashboard to visualize the impact of removing forecasted bottlenecks. - Derived 38% efficiency in preventing late payments made to vendors using A/B Testing, which was the result of better forecasting from AI models on bottlenecks and explained AI decisions to business with LIME and SHAP tools. - Automated data engineering pipelines (ETL, SQL) for data transformations needed to support AI models using Google Dataproc, with compliance to finance data auditing and region-based data usage policies. - Scaled AI forecasting solution to different stakeholder regions using Google Cloud based Kubernetes, Docker and Terraform scripts, CI/CD with GitHub actions, enabling stakeholders to reduce late payment losses across organization. - Contributed to campaign analysis for shell downstream - lubricants, with interactive dashboards (PowerBI) that enabled stakeholders to get marketing campaign insights and metrics from various data sources. - Enabled faster calculation of campaign metrics (CTR, ROAS, impressions, conversions) for dashboards by approximate 40% with integration of real-time data services managed with Kafka

Machine Learning Engineer at HTC Global Services, Mar 2019 - May 2022 · 3 yrs 3 mos, Guindy, Chennai - India, - Trained and productionized an image segmentation model (PyTorch, Deep Learning) to automate the assessment of vehicle damage in auto-insurance claims, reducing processing times and cost by 35% - Collaborated with multiple teams for exposing AI models to end-users via Flask REST API’s, enabling users to get instant access to their insurance quotes, model features and improve automation efficiency for insurance clients via mobile/web. - Enabled faster user responses by moving model to CUDA, improving user mobile application experience on scale. - Deployed the automation solution for client users to consume over the internet with AWS cloud deployment (Docker, Kubernetes), hosting the AI model using sagemaker, monitored with cloudwatch enabling services with no downtime. - Developed a chatbot assistant to give employees under the organization context relevant information on organizational policies, replacing third party services for chatbot systems which reduced 25% cost to company. - Streamlined data pipelines needed for assistants NLP model (BERT - Hugging Face) using Spark distributed services, which enabled larger data handling, better system processes improving reliability when data is stored back to Google cloud storage.

Education

Master of Science - MS, Computer Information systems (Data Science & AI), California State University, Long Beach

Bachelor of Engineering - BE, Mechanical Engineering, SRM IST Chennai