FOUNDER Mar 2026 — Present
Founder & Principal Engineer
Open-core LLM gateway + AI-agent reliability platform. Observability, guardrails, audit trails — catching agent failures before execution.
I keep critical systems alive.
Now I build the tools that keep AI agents alive.
Fifteen years running mission-critical financial systems — currently Second Vice President at Northern Trust. Founder of Tracelane, an open-core LLM gateway and AI-agent observability platform. I bring an operator’s paranoia to AI infrastructure: assume failure, trace everything, catch it before it executes.
15yrs
Mission-critical production ops
99%
SLA uptime, institutional platforms
4.6ms
Tracelane p99 gateway overhead
375µs
p99 inline guardrail dispatch
// currently
OPERATOR Oct 2021 — Present
Second VP, Application Support Consultant
Northern Trust
Senior escalation point for institutional financial platforms across EMEA/APAC. 99% SLA. SOC 2 Type II audits with KPMG. 12+ engineers mentored.
// founder work
Open-core · LLM gateway + AI-agent observability
tracelane.dev ↗
Reliability infrastructure for AI agents.
4.6 ms p99 gateway overhead
375 µs p99 guardrail dispatch
11.5k rows/sec ingest, single node
One platform for routing, tracing and governing production agents: an LLM gateway with provider failover, OpenTelemetry GenAI traces over ClickHouse, inline guardrails, MCP failure-signature detection, and cryptographic audit trails. Built solo, in Rust and TypeScript, while running a day job in institutional finance.
- Multi-provider routing · failover · queue replay
- OpenTelemetry GenAI traces — LLM, tool, browser, MCP
- Inline guardrails + eval-gated CI
- SSO/RBAC · BYOK · tenant quotas · audit trails
// principles
How I operate
Fifteen years of production incidents leave you with convictions. These are mine.
Assume failure
Every system fails; the question is whether you find out from a trace or from a customer. Design for the failure first — observability, guardrails and rollback are the product, not the afterthought.
Be customer zero
I run Tracelane on Tracelane. Production use exposes the failures that automated tests never catch — that feedback loop is worth more than any roadmap.
Boring reliability, novel systems
AI agents are new; the discipline of keeping them alive is not. SLOs, runbooks, DR drills and post-incident rigor transfer directly from institutional finance to agent infrastructure.
// track record
Fifteen years, zero abandoned systems.
Founder & Principal Engineer
Tracelane
Architected and launched an open-core AI-agent reliability platform: LLM gateway, OTel GenAI tracing, ClickHouse analytics, guardrails, MCP failure detection. Eval-gated CI, blue-green deploys, provider failover. Used it as customer zero to find failures tests never caught.
Second VP, Application Support Consultant
Northern Trust
L2 lead for mission-critical fund-servicing platforms, EMEA/APAC. 99% SLA. Led Bedrock monitoring initiative end-to-end. Annual EMEA/US disaster-recovery exercises to RTO/RPO. Hired and mentored 7 DevOps engineers through Kubernetes transformation. Cut recurring incidents ~30% and manual ops ~30%. KPMG / SOC 2 Type II audit partner.
Technology Lead, L2 Application Support
Northern Trust · via Infosys
Managed 12+ specialists across 20+ global fund-servicing applications. Zero-downtime production upgrades. ~30% recurring-incident reduction via trend analysis and automation.
Technology Analyst, Application Support
NatWest Group · via Infosys
24/7 production support for trade affirmation and confirmation platforms. Live SQL fixes for customer-impacting incidents. ServiceNow / JIRA incident and change management.
Software Engineer, ETL Support
AIG · via Mphasis
Oracle data-warehouse ETL, PL/SQL development, SQL performance tuning, business-critical BFSI reporting.
B.E. Electronics & Communication — Anna University, Chennai · 2007 — 2011 · First Class with Distinction · ITIL V3 Foundation · Agile Kanban
// stack
AI infrastructure
Engineering
Operations
// also running