Palantir - Full Time

Palantir - Full Time

After joining full-time, I loved that I was able to return back to my former team, but work on several different problems unlike I've ever seen before. I was a backend developer and it was super nice.

While I've always felt that I've learnt a lot in my internships, it doesn't compare to this full time experience. Here, I got to really hone my system architecture, low-latency design, clean choices, and intuition for how things should work. (Thank you to my team for keeping extremely high code standards! 😁)

Overall, it is great though! I am happy I was able to not just learn myself, but also be able to help onboard others to the team and share what I knew.

Here's a few notable achievements from my time there (so far! I'm currently there):

  • Diagnosed duplicate execution and database contention, then independently built distributed concurrency libraries with service-scoped work ownership, optimizing deletion by 10x+.
  • Designed heartbeat-based leases, TTL expiration, and automated work reclamation for node-failure recovery, owning architecture and rollout across shared libraries and services without downstream integration changes.
  • Designed and built a Java security-policy service leveraging just-in-time prefix tries, reducing evaluation time for comparable requests from ~3 minutes to ~200 ms (~900x faster).
  • Developed an agentic knowledge marketplace spanning incident ingestion, skill synthesis, updates, conflict detection, and pruning, enabling a continuously maintained contextual-knowledge base used across agentic workflows and standalone agents, reducing token spend by 15%.
  • Partnered with security stakeholders to deliver a compliant self-service passkey recovery solution with identity verification and auditable event logs, reducing recovery turnaround from weeks to minutes - user wait times from 7 days to 2 minutes - cutting support overhead and saving ~10,000 employee hours weekly. (I loved this project!)
  • Developed an AI-powered internal assistant using a subagent architecture and MCP servers to unify Sourcegraph, Grafana, Jira, and other monitoring systems, automating root-cause analysis and accelerating incident resolution. (Metrics are a bit unknown as of writing this)
  • Mentored two engineers through structured onboarding, technical guidance, and design reviews, enabling both to join the team's support rotation within 1.5 months.

Enjoying it so far!