Selected work

Some of the systems I've designed or led, and what each one changed.

A customer-support AI agent, and the MCP servers behind it

Okta, 2025 to 2026. LLMs and agents.

Tech lead for Okta’s customer-support GenAI agent, including the chat widget on support.okta.com and the search behind it.

Built
The agent itself, taken through security architecture review, threat modeling, and data-classification controls before it went public. Alongside it, MCP servers and multi-agent workflows that let agents work over enterprise systems.
Stack
  • Python
  • LangGraph
  • AWS
  • MCP
  • RAG

An event-driven integration platform for the revenue CRM

LinkedIn, 2022 to 2025. Data systems.

Authored the RFCs and led the platform that keeps Salesforce, Dynamics 365, and the sales tools around them in sync, built on Kafka and change data capture.

Result
More than 1 million transactions a day. Sync latency went from one hour to near real time.
Stack
  • Java
  • Kafka
  • Change data capture
  • Salesforce
  • Dynamics 365

Data pipelines and integrations around the CRM

LinkedIn, 2017 to 2025. Data systems.

Data engineering alongside the CRM work from 2017: scheduled batch pipelines on Azkaban, and the integrations that move data between the CRM and the systems around it.

Stack
  • Azkaban
  • Batch pipelines
  • Integrations

A territory assignment engine

LinkedIn, 2022 to 2025. Data systems.

The engine that assigns sales territories each planning cycle, processing more than 5 million records per cycle.

Result
Planning cycle cut to 11.5 hours. Error rate from 30% to 0.2%.

A GPT-based model inside the CRM

LinkedIn, 2022 to 2025. LLMs and agents.

Integrated LinkedIn’s GPT-based model into Salesforce for the sales organization.

Result
A 13% productivity gain for more than 10,000 users.

Quote-to-cash on Salesforce CPQ

LinkedIn, 2017 to 2022. Data systems.

Tech lead for the CPQ program behind LinkedIn’s global sales organization: the quote calculator, product and price rules, approvals, and the sync from quotes to opportunities.

Result
Quote processing went from three days to real time, unblocking $350M a year in deal flow.
Stack
  • Salesforce CPQ
  • Apex
  • JavaScript
  • Java
  • Kafka

Managing the CRM engineering team

LinkedIn, 2020 to 2025. GTM systems.

Managed LinkedIn’s CRM engineering team alongside the architecture work: performance reviews, promotions, planning, and roadmaps presented to executives. Partnered with Microsoft on the Dynamics 365 program.

Earlier: payments at Pandora, enterprise CRM in Pune

Pandora, Adobe, ZS, HCL, Cybage, 2010 to 2017. Data systems.

A payments portal and subscription systems for Pandora, Java services at Adobe, and before that five years of enterprise CRM engineering in Pune: lead routing, approvals, integrations, and single sign-on. The Salesforce posts in the archive come from these years.