Wake Up Daily As If Every Day Were SIGMOD

I had two choices for spending my discretionary income this year: Tomorrowland Thailand or SIGMOD Bangalore.

During the week of the conference, I would reach home by midnight after post-dinner conversations with database researchers I had long wanted to meet. Then I’d wake up at 4 a.m. to prepare for the day’s sessions and scheduled chats. I cherished every minute of the conference because I knew I would probably never get this kind of access again—access to the people whose papers I had read, whose lectures I had watched on YouTube, and whose tweets I had bookmarked.

Notes to myself, from these conversations:

1. Systems, Not Just Databases

Don’t narrow yourself to data processing. Read widely beyond SIGMOD/VLDB/EDBT/ICDE. Build a systems toolbox—reverse-engineer the primitives that papers use (caching, predicate pushdown, etc.). Good research applies the right tool to the right problem; great research expands the toolbox.

2. Pull the Trigger

If you want to do a PhD, commit now. Grit, patience, and stress tolerance matter more than raw brilliance.

3. Leverage Your ML Background

Use it as a differentiator when trying to break into systems graduate school.

4. Pick the Right Advisor

Early-career professors are hungry to publish fast at top venues; senior professors have stronger networks to embed you into.

5. Own Your Work

Use your advisor as a resource to unblock you, but you alone are accountable for progress.

6. Coherence Over Collection

Your research projects should tell a coherent story, not scatter in random directions. Pursue projects that are intellectually interesting to you and “useful” to the world.

7. Abstract Problems Down to Their Core

Learn to strip problems down until you reach their essence.

8. Mini Milestones & Rewards

Create them yourself. Clear, immediate feedback is rare in research. When stuck, talk to people sooner.

9. Market Your Work

Be ambitious. Craft memorable paper titles that sell the idea while sounding scientifically credible. Simplify your slides, rehearse your talks, and modulate your voice.

10. Be Fun to Work With

Be likable, don’t gossip.

11. Microbenchmarking

Get good at writing tight microbenchmarks to validate hunches. Hand-write unit tests for correctness; integration tests for performance.

12. Theory Papers

When reading PODS-style work, ignore the notation at first and follow the examples through.