About
DineshKumar Sarangapani
Summary
I am an AI Architect. By company role, I serve as a Distinguished Engineer in AI at Trimble, setting the technical direction for how AI is built into enterprise products and leading the engineering behind the shared platform that makes that possible.
I got here by building the systems other teams depend on. Before AI, that meant identity, authorization, and cloud platforms for the whole company. On our AI platform, I architected the multi-cloud model gateway and the enterprise MCP gateway from the ground up. Together, they have carried more than 130 billion tokens and 110,000 tool calls in production across 1,000+ weekly active agents. I also scaled our cloud compute on Amazon EKS with Karpenter Spot provisioning, built a unified OpenTelemetry pipeline into ClickHouse, and achieved ISO/IEC 27001, SOC 2 Type II, and ISO/IEC 42001 compliance as code. The lesson is simple: get the foundation right, and many teams can move fast on top of it.
I lead through architecture and standards more than through org charts. I write decisions down, review designs early, and help engineers grow into owners of the systems they build. A defining strength—and sometimes a quiet weakness—is the ability to foresee a distant future and build for it early. When you build the foundation years before the storm arrives, people rarely realize what was saved, because the crisis never happened. But that quiet foresight is what keeps platforms standing when scale explodes.
15+ years in production systems taught me where AI breaks. I find it before your users do.
Where life stands in 2026: milestones, scale & systems
A dedicated reflection on carrying 130B+ tokens, EKS Spot compute, and ISO 42001 AI governance.
Principles
How I work
Platforms before features
Solve a hard problem once, in a shared place, so every team benefits from it.
Measure before you scale
If we cannot measure the quality of an AI feature, it is not ready to ship widely.
Make the safe path the easy path
Security and responsible AI work best as defaults that teams get for free, not as checklists.
Build for the unseen future
My strength is seeing around corners and building for a distant future before others realize it is needed. It is also a quiet weakness: when you solve a crisis years ahead of time, nobody notices the disaster that never arrived.
Write decisions down
A short written record beats a long meeting, and it outlives the people in the room.
Stay close to the code
I still build. It keeps my designs honest and my reviews useful.
Career
Experience
-
2025 – Present
Distinguished Engineer, AI · Trimble
Lead architect for the enterprise Agentic AI Platform, setting technical direction and building the shared infrastructure.
- Architected the multi-cloud model gateway: dynamic routing, circuit breakers, and cross-cloud failover across 60+ models from four providers, serving 130B+ tokens in production
- Built the enterprise Model Context Protocol (MCP) tool broker: automated OpenAPI-to-MCP translation with JSON schema dereferencing, handling 110K+ tool calls across 120+ servers
- Engineered the Token Vault for delegated On-Behalf-Of (OBO) user authorization, allowing agents to access third-party tools securely without prompt leakage
- Migrated compute to Amazon EKS on EC2 Auto Mode with Karpenter Spot autoscaling, cutting batch document processing and ingestion costs by 70–90%
- Architected the unified OpenTelemetry pipeline: forking high-cardinality GenAI traces into real-time APM, financial token ledgers, and ClickHouse OLAP
- Achieved ISO/IEC 27001, SOC 2 Type II, and ISO/IEC 42001 (AI Management System) compliance with automated posture auditing and continuous evidence as code
- Built self-service IssueOps and GitOps deployment automation, enabling developers and AI agents to safely ship compliant services in minutes
-
2022 – 2024
Distinguished Engineer, Cloud & AI · Trimble
Architected the cloud platform for the transportation business and led early machine learning and deep learning initiatives.
- Architected the multi-tenant cloud platform for the transportation business, modernizing core services and data pipelines
- Led early machine learning and deep learning initiatives, introducing neural models for domain predictions and classification
- Shipped AI-powered assistants for customer support and operations teams across chat and voice channels
- Evaluated early foundation models and established safe AI development practices across engineering teams
- Led architecture reviews and mentored senior engineers across multiple product lines
-
2020 – 2022
Principal Software Engineer · Trimble
Led platform engineering for shared services used across the company.
- Built a standards-based (OIDC) identity platform serving 30M+ users and 300+ apps at 99.99% uptime
- Designed an authorization service adopted by 40+ applications
- Built operations support assistants across Google Chat, Slack, and phone
- Built event sync between services on Kafka and SQS
-
2015 – 2020
Senior Software Engineer · Trimble
Built core platform services and the cloud foundations under them.
- Built a shared coordination service used by other platform services
- Delivered serverless applications and Azure AD B2C identity integrations
- Applied the AWS Well-Architected Framework across services
- Brought a secure development lifecycle into the regular delivery process
-
2013 – 2015
Senior Software Engineer · Lister Technologies
Designed enterprise applications and led technical delivery for client projects.
-
2010 – 2013
Software Engineer · Aspire Systems
Built and maintained enterprise software across the full development lifecycle.
Toolbox
Skills
AI and Agentic Systems
- AI agents and multi-agent systems
- Model Context Protocol (MCP)
- Multi-cloud LLM gateways
- AI evaluation and benchmarks
- Responsible AI and guardrails
- Token Vault and agent OAuth
Platform Architecture
- Distributed systems
- Streaming APIs (SSE, chunked)
- Identity and OIDC federation
- High-throughput microservices
- Enterprise multi-tenancy
Cloud and Kubernetes
- AWS
- Azure
- Kubernetes
- Karpenter
- Terraform
- Docker
Data and Observability
- OpenTelemetry (OTEL)
- ClickHouse OLAP
- Amazon ElastiCache Redis
- MongoDB Atlas
- PostgreSQL
- Kafka
Engineering and Languages
- Python
- FastAPI
- TypeScript
- Node.js
- Java
- Spring Boot
Governance and Delivery
- ISO/IEC 42001 (AIMS)
- ISO/IEC 27001:2022
- SOC 2 Type II
- IssueOps and GitOps (ArgoCD)
- Automated posture audit as code
Learning
Education
-
Indian Institute of Technology Madras
BS, Data Science and Applications, 2021 – 2023
-
Birla Institute of Technology and Science, Pilani
MS, Computer Science, 2012 – 2015
-
Anna University
BE, Computer Science and Engineering, 2006 – 2010
Learning
Certifications
-
Deep Learning Specialization
DeepLearning.AI, Coursera. Five courses.
Neural Networks and Deep Learning Improving Deep Neural Networks Structuring Machine Learning Projects Convolutional Neural Networks Sequence Models
-
Machine Learning
Stanford University, Coursera
-
Introduction to Programming Using Python
Coursera