Skip to content
DineshKumar Sarangapani

AI Architect · Distinguished Engineer, AI at Trimble

DineshKumar Sarangapani

I set the technical direction for how AI is built into enterprise products, and I lead the platform that makes it possible.

15+ years in production systems taught me where AI breaks. I find it before your users do.

For more than 15 years I have built the shared systems many products depend on. On our AI platform I built the model gateway and the MCP gateway from the ground up, and they carry the platform's model and tool traffic. Today I work on the hard parts of enterprise AI: one architecture for many teams, safe use of models and company data, and a clear bar for what is ready to ship.

About me LinkedIn
Portrait of DineshKumar Sarangapani
tokens through the model gateway
130B+
tool calls through the MCP gateway
110K+
models from four providers
60+
MCP servers connected
120+
agents active each week
1,000+
years building platforms
15+

Approximate, rounded figures that show the scale of my work.

2026 In Review

Where life stands in 2026: milestones, scale & systems

A retrospective on carrying 130B+ tokens, architecting multi-cloud gateways, scaling Kubernetes with Spot compute, and leading through code.

View 2026 Milestones

Focus

What I am responsible for

Architecture

One platform shape that many product teams can build on. I design the shared foundation so each team ships AI features without rebuilding the hard parts.

Trust

Safe use of models and company data, by design. Access, data boundaries, and guardrails are part of the architecture from day one.

Quality bar

How teams decide an AI feature is good enough to ship. I push for measurement before scale, so quality is a number and not an opinion.

Leadership

The standards, reviews, and people around the platform. I write decisions down, review designs early, and grow engineers into owners.

Track record

Selected outcomes

  • Built the model gateway from the ground up

    One governed entry point to 60+ models from four providers. It has served 130B+ tokens in production with cross-cloud failover, circuit breakers, and latency telemetry.

  • Built the MCP gateway from the ground up

    Secure, audited access from agents to 120+ MCP servers. Handled 110K+ tool calls with dynamic OpenAPI translation, schema dereferencing, and Token Vault identity.

  • Compute scale on EKS with Karpenter Spot autoscaling

    Migrated compute to Amazon EKS with Karpenter Spot provisioning, cutting batch processing costs by 70–90% while scaling to thousands of concurrent worker pods.

  • Deploy in minutes, compliant by default

    Built IssueOps and GitOps automation where engineers and AI agents safely ship services. Achieved ISO/IEC 27001, SOC 2, and ISO/IEC 42001 compliance as code.

  • Shared platforms before AI

    Built company-wide identity serving 30M+ users and an authorization service adopted by 40+ applications. The same foundations now sit under the AI platform.

Writing

Selected writing

View all writing

Other articles

Contact

Let's talk

Open to conversations on enterprise AI architecture, platform strategy, and engineering leadership.

EmailLinkedInGitHub