Saai Sudarsanan
Senior Software Engineer I at Swym Corporation. Three years of Kubernetes, observability, and platform engineering — now building AI ops tooling. Pursuing OMSCS at Georgia Tech.
View full detailed timeline →The Arc
Intern → Senior SWE I in 3 yearsKey Milestones
Click any card to expandAlfred — Internal AI Ops Assistant
Built Swym's AI operations assistant from scratch in ~9 days. Lives in Slack, thinks with Gemini.
BFCM 2025 — 1.01M Peak RPM
Handled Black Friday / Cyber Monday peak traffic of 1.01M RPM on Kubernetes. Zero incidents.
Self-Serve Monitoring Platform
Built a fully self-serve, multi-team monitoring platform using Thanos, kube-prometheus-stack, and custom exporters.
Full Infrastructure Modernization — 3 New AKS Clusters
Designed and migrated all Swym workloads to 3 new AKS clusters — prod, staging, ops — built from scratch.
AG4C Migration — $120/day Saved
Migrated swym-store (all tiers) to Azure Application Gateway for Containers, cutting $120/day in infra costs.
MCP-Gateway for Observability
Unified MCP server exposing Jaeger, OpenSearch, and Prometheus as one AI-accessible endpoint with PII redaction.
Skills
By domainInfrastructure & Orchestration
Observability & Reliability
Cloud & Networking
AI & Automation
CI/CD & Development
Security & Compliance
Education
Master of Science, Computer Science (OMSCS)
Georgia Institute of Technology
B.Tech, Information Technology
SASTRA Deemed University
Notable Projects
Alfred
Internal AI ops assistant for Swym — Gemini 2.5 Flash agentic loop with MCP tool access to production Kubernetes, Prometheus, Jaeger, OpenSearch, and GitHub. Thread-level and long-term memory, human approval gates, proactive idle research. Built in ~9 days. Deployed on Kubernetes.
MCP-Gateway
Unified FastMCP gateway exposing Jaeger, OpenSearch, and Prometheus as one AI tool endpoint. Supports namespace filtering and includes PII redaction middleware before responses reach the LLM.
Benchmarking suite for Clojure logging frameworks. Identified significant CPU overhead from pformat pretty-printing and proved the fix — reducing log size and JVM GC pressure in production.
turinglib
Lightweight Python library for building and simulating Turing Machines. Object-oriented interface for defining tape symbols, actions, states, and state machines.
Certifications
Writing
22 articles across Medium, LinkedIn & SubstackSetup Jaeger Operator with Opensearch for Kubernetes
A hands-on guide for integrating tracing and observability using Jaeger and OpenSearch in a K8s cluster.
Medium
What are Convolutional Neural Networks?
Breaking down CNNs — the architecture powering modern computer vision.
Medium
Gradient Descent from Scratch
Building the Gradient Descent Algorithm from scratch and explaining it with Manim
Medium