Naman Gupta

Backend on a multi-tenant credit platform running 10K+ TPS and 100M+ transactions a month at 99.99% availability. Own the Credit Card Management System and the REST/SOA layers that standardize responses across core transaction processors and third-party backends. Architected a reporting service end-to-end — system design, data modeling, implementation — plus the pipeline that ingests client CSVs over SFTP through Chronos and Kafka into Postgres and S3. Kafka reconciliation workflows hold consistency under concurrent load via message ordering and account-level locking. Drove initial request latency down ~80% to 30ms.

A logging system for LLM calls, built as four pieces that only make sense together: a Python SDK that wraps any provider client without blocking it and masks PII before anything leaves the process; a Django 5 / Django Ninja ingestion API taking batched envelopes; Kafka partitioned by session id with manual commits and a dead-letter topic; and a React dashboard reading Postgres — latency percentiles, cost by model, token usage, error bursts, plus an SSE live tail straight off the topic. The framing was that an inference log is evidence: it has to survive the process that produced it. Runs on one docker compose command, seeded with a week of telemetry so the dashboards have something to show immediately.

Built and validated 35+ predictive signals on US and China equity markets. Discovered alphas with Sharpe ratios from 1.25 to 2.91, turnover up to 102%, cross-correlation under 0.7. Top 0.5% globally (rank 26) in AlphaVerse Premier League and global rank 302 in International Quant Championship.