Alvin Tan

Engineer with 11+ years at Google and Microsoft building backend systems, product infrastructure, and user-facing product features from initial design through launch and scale. I stayed on the individual contributor track throughout, focused on solving hard technical problems and shipping rather than optimizing for promotion cycles. Strongest in workflow orchestration, latency reduction, data correctness, and AI-assisted product behavior.

Best used where the job is to build the system, ship the fix, and make it hold up in production.

Software Engineer III at Google (2021-10 – 2025-12)

Worked across Google Labs Stitch, Google Vids, and Looker Studio on AI-assisted product tooling, backend workflows, and migration systems. Stitch was the most product-facing of the three, centered on iterative UI refinement and exporting React and Next.js code engineers could actually use.

• Built the versioned UI graph model for Google Labs Stitch that reduced unintended code changes across prompt refinements by 60–70%.
• Improved Stitch's exported React and Next.js output by preserving component identity across iterations, making the generated code more usable in real codebases.
• Redesigned Google Vids' export backend ahead of launch, cutting export failures from 7–10% to below 0.5% and preview latency from 75–120 seconds to 8–12 seconds.
• Reduced duplicate rendering work in Google Vids by introducing deterministic job identifiers for workflow stages.
• Maintained interactive preview performance during enterprise export spikes by partitioning worker capacity with priority queues.
• Built a replayable migration model for Looker Studio that reduced silent corruption risk during the Data Studio transition.

Software Engineer II at Google (2017-11 – 2021-10)

Worked across Google Keep, Google Fit, and Google Lens on backend performance, data correctness, privacy-sensitive systems, and inference routing.

• Fixed a write-path bottleneck in Google Keep by replacing synchronous transactions with async Pub/Sub reconciliation and redesigning PostgreSQL indexes, cutting p95 mutation latency from 4.2 seconds to under 400ms.
• Improved Google Keep AI-assisted creation success rates by more than 30% by removing Gemini inference from the synchronous request path.
• Built a normalization gateway for Google Fit that eliminated silent health data corruption across third-party Health Connect providers after rollout.
• Reworked Google Fit permission revocation handling to close a privacy compliance gap where revoked consent was not enforced for up to 15 minutes after withdrawal.
• Improved Google Lens OCR responsiveness by routing expensive OCR only to frames most likely to improve output quality.
• Redesigned Google Keep search for scale by replacing fan-out joins with read-optimized PostgreSQL projection tables.

Software Engineer at Microsoft (2014-06 – 2017-11)

Worked across Cortana, Microsoft Sway, Bing, and Microsoft Health on backend orchestration, frontend state handling, API contracts, and device-data processing for products operating at large user scale and high delivery pressure.

• Redesigned Cortana's service orchestration before and through the Windows 10 launch, cutting timeout-driven failures by roughly 60% and median latency by 22% under real user traffic.
• Fixed Microsoft Sway's layout divergence problem, reducing visible layout corrections from more than 30% of active-editor page loads to under 3%.
• Improved Bing dense-query aggregation latency by 42% in a one-week production A/B experiment by parallelizing 11 backend source calls across mapped dependency stages.
• Built a validation layer for Microsoft Band health data that caught firmware-specific corruption issues before they silently propagated.
• Standardized API payload contracts across Bing, Cortana, and Sway, reducing integration regressions caused by contract mismatches.

Bachelor of Science in Computer Science – University of California, Santa Barbara (2010 – 2014)