$ whoami

Sanchit Gupta
Software Engineer — AI Infrastructure & Backend Systems

$ ./deploy_history.sh --last 4

✓ Migrated inference stack to vLLM        → +3.0x throughput
✓ Rewrote hot-path caching layer          → -40% p99 latency
✓ Stood up on-prem GPU deployment         → zero external data egress
✓ auth + rate-limiting layer      → sustained 0 downtime incidents

$ cat tech-stack.txt 

• Python • FastAPI • Docker • PostgreSQL
• Redis • vLLM • NGINX • Linux • Git • GitHub
• Prometheus • Grafana • Monitoring

$ docker logs api-gateway --since 10m | grep ERROR

2026-07-12T09:14:02Z ERROR db.pool: connection timeout after 5000ms
2026-07-12T09:14:02Z ERROR db.pool: retrying (attempt 1/3)

$ docker exec -it postgres bash
root@postgres:/# psql -U user -d aiaas

aiaas=> SELECT count(*) FROM pg_stat_activity;
 count
-------
   187
(1 row)

aiaas=> \q
root@postgres:/# exit

$ redis-cli INFO clients | grep connected_clients

connected_clients:342

$ pytest tests/ -k "rate_limit" -v

tests/test_middleware.py::test_rate_limit_burst PASSED
tests/test_middleware.py::test_rate_limit_reset PASSED
tests/test_middleware.py::test_rate_limit_per_ip PASSED

3 passed in 1.42s

$ git commit -am "fix: increase db pool size to handle traffic spike"

[main 7c2e1a9] fix: increase db pool size to handle traffic spike
 1 file changed, 2 insertions(+), 1 deletion(-)

$ git log --oneline -5 --grep="perf"

a3f9c1e perf: batch inference requests, +18% throughput
7bd2e40 perf: connection pooling for postgres, -12ms avg
c910af3 perf: redis pipeline for hot-key lookups
1e6a8b7 perf: async request queue, reduce lock contention
f402d9c perf: precompute embeddings on write path

$ docker build -t registry.internal/api:v2.4.1 .

[+] Building 42.3s (14/14) FINISHED
 => exporting to image                                    0.4s
 => => writing image sha256:9f2a1c...                     0.0s
 => => naming to registry.internal/api:v2.4.1              0.0s

$ ./run_tests.sh --coverage

Ran 214 tests in 38.7s
Coverage: 91%
All tests passed ✓

$ kubectl rollout status deployment/api-gateway -n prod

Waiting for rollout to finish: 2 of 3 updated replicas are available...
deployment "api-gateway" successfully rolled out

$ cat postmortems/*.md | grep -c "root cause"

0   # no unresolved incidents on record