Subash B

AI Engineer with 4+ years across AI, automation, and enterprise software, with strong hands-on experience building production-grade LLM applications for global enterprise clients in pharma and logistics. Core contributor to Takeda Predictive Insights Centre, a GenAI chatbot serving 500+ stakeholders, and currently contributing to its migration into an MCP-based agentic architecture. Skilled in LLM function calling, API/tool orchestration, prompt engineering, NL-to-SQL, RAG, semantic search, and production backend integration.

Senior Engineer - AI - Altimetrik - Chennai, India

(2026-06)

Takeda Predictive Insights Center - MCP-Based Agentic Architecture Migration

• Key developer on the migration of the Takeda enterprise chatbot from a custom LLM orchestration layer to a Model Context Protocol (MCP) based agentic system using ChatKit.
• Defined MCP tool specifications that wrap REST and GraphQL APIs as discoverable tools, enabling LLM-driven tool selection, parameter filling, execution, and natural language response synthesis.
• Contributed to a domain-partitioned service architecture across logistics, manufacturing, and transportation MCP services to improve query routing precision and reduce response latency.
• Re-levelled to Senior Engineer - AI in Jun 2026 under company career framework; prior promotion from Associate Engineer to Engineer in Jan 2026.

Engineer - Data and AI - Altimetrik - Chennai, India

(2026-01 - 2026-06)

Takeda Predictive Insights Center - Enterprise LLM Chatbot

• Developed a production AI chatbot enabling 500+ Takeda stakeholders across logistics, manufacturing, and transportation to query complex operational data in natural language across dev2, test2, dev, test, and prod environments.
• Reduced manual operational data lookup from hours or days to seconds by enabling natural-language access through LLM-driven API orchestration.
• Engineered LLM function-calling pipelines using OpenAI GPT models, evolved from earlier AWS Bedrock/Claude-based implementations, to identify APIs, extract parameters, and invoke REST/GraphQL endpoints.
• Built a resilient NL-to-SQL fallback where the LLM constructs validated PostgreSQL queries using schema context stored in AWS S3 when APIs are unavailable.
• Supported validation of LLM outputs, API selection, parameter extraction, and NL-to-SQL fallback responses to improve reliability across production chatbot workflows.
• Owned knowledge base design for API specifications, parameter schemas, and table definitions in AWS S3, and implemented context injection strategies by query domain.
• Contributed to FastAPI backend development, including GraphQL chatbot endpoints and Okta-based API authorization flows.

Associate Software Engineer - Data and AI - Altimetrik - Chennai, India

(2022-12 - 2025-12)

Takeda Predictive Insights Center

• Joined the Takeda PIC project and contributed to the production LLM chatbot lifecycle, including API/tool orchestration, knowledge base refinement, and enterprise stakeholder query support.
• Built an internal document Q&A chatbot over 200+ HR policy documents using OpenAI embeddings and FAISS for semantic retrieval.
• Helped the HR chatbot gain internal recognition and later implementation in the company's official portal.

Project Intern - AI and Automation - Altimetrik - Chennai, India

(2022-01 - 2022-12)

• Automated QA workflows for Citi Bank using Java, Selenium, and Jenkins CI/CD, receiving a Kudos Award for exceptional delivery.

B.E. - Computer Science and Engineering - Dhanalakshmi College of Engineering, Anna University (2018 - 2022)