Hiring profile
I own backend and data infrastructure systems that scale to 200+ customers under production constraints. Rearchitected data ingestion from legacy cron-based batch (days-late) to real-time CDC with sub-5-minute freshness for webhook-capable platforms, up to 6 hours for others.
I work best owning backend and data infrastructure systems end-to-end. I'm drawn to roles where I can influence architecture across multiple teams, lead small teams through scaling, and own both technical direction and operational reliability.
Relevant role fit: Staff Backend Engineer · Staff Platform Engineer · Principal Engineer · Founding Engineer · Senior Backend Engineer · Senior Platform Engineer · Senior Data Infrastructure Engineer
Cost optimization · data infrastructure
77.8% cost reduction (6× more efficient). Migrated customer segmentation from BigQuery Views to ClickHouse using batched deterministic hashing (cityHash64) and right-sized compute (64GB → 16GB RAM). The workload was wrong for the database model, not a tuning problem.
₹1.4 lakhs/month savings · 6× efficiency gainOwnership · system · scale
I owned the rebuild of customer and order data infrastructure serving 200+ merchants across five commerce platforms across five commerce platforms. Freshness ranges from sub-5-minute for webhook-capable platforms to ~6 hours for legacy integrations, up from days-late baseline.
Sub-5-minute to 6-hour tiered freshnessOwnership · product · adoption
I initiated and built a three-tier recommendation product, from trust-first manual blocks to live ranking across five regions.
80%+ of merchants who trialled Smart tier adopted itOwnership · data product · rollout
I owned the query model and builder that unified four source systems for Customer Success teams.
1,000+ active segments within six months of launchStart with the data platform, then read the production incident. These show architecture ownership and operational judgment fastest.
Primary case study · recommended first read
I owned architecture, migration, and rollout for customer and order data across 200+ merchants on five commerce platforms. Freshness: sub-5-minute for webhooks, ~1–6 hours for others (vs. days-late baseline).
Production incident · technical deep dive
Debezium consumed 60% CPU. I stopped the snapshot, used the existing Datastream path to recover, and changed the rollout strategy.
A decade of production lessons: query shape, lock contention, silent data failure, memory pressure, database constraints, resilience patterns, and the judgment to trade freshness for stability when the system needs it.
Reversed join-before-limit pagination and cut a 3+ minute query to under 500ms.
Isolated a post-migration regression across 100+ merchants and restored baseline by reducing load parallelism.
Replaced an unfit full join with deterministic bucketed hashing: fixed memory, correct diffs, 6× lower cycle cost.
Turned daily Pub/Sub consumer OOM crashes into a pause/resume pattern adopted across consuming services.
Also shipped Affairs Map (LLM production pipeline), Fruggy (Flutter, 1K+ downloads), and a few other independent projects — happy to share the full list on request.
Download my resume and tell me what you're building.