Skip to content
Experience

Experience: platform, infrastructure and AI

Two tracks running at the same time: a full-time engineering role on a multi-cluster Kubernetes product, and client work that has taken me through GitOps automation, backend services and, more recently, real-time AI systems.

Senior Software Engineer

Jan 2023 – Present

AppsCode Inc.

Software Engineer, Jan 2023 to Dec 2024. Promoted to Senior Software Engineer, Jan 2025.

  • Built a centralized multi-cluster management solution for public and private Kubernetes clusters via the ACE Dashboard.
  • Developed Kubernetes operators and APIs to automate authentication and authorization with secure kubeconfig generation.
  • Designed and implemented a multi-tenant monitoring system using OpenTelemetry, Thanos, ClickHouse, and Perses.
  • Packaged ACE as a self-hosted service for deployment across vendor-managed and self-managed Kubernetes clusters.
  • Integrated high-impact features into the ACE API Server, enhancing platform capabilities and scalability.
  • Contributed to the deployment, maintenance and reliability of AppsCode production servers.

Senior Software Engineer (Top Rated Plus)

2021 – Present

Global Clients (Upwork)

  • Built a real-time voice AI agent: streaming speech-to-text, access-filtered hybrid retrieval over Qdrant, LLM generation and cloned-voice synthesis, delivering spoken replies 0.83–1.17s after end of turn, plus a lip-synced video avatar.
  • Designed the GPU serving layer behind it: pod-first routing with serverless overflow, per-endpoint circuit breakers, and an autoscaler that learns each pod's real concurrency, and benchmarked capacity at ~11.5 rps and ~25 concurrent users per pod at p95 ≤ 6s.
  • Built the voice backend for a Unity game with an LLM-powered companion character, cutting time-to-first-audio from ~7s to ~1.5s and removing 40-60% of inference calls with an event pre-classifier.
  • Delivered an infrastructure automation suite for enterprise OpenShift clusters using Ansible and ArgoCD for GitOps-driven continuous deployment.
  • Architected and deployed highly scalable backend systems for international clients using Go and Python.
  • Automated infrastructure provisioning and multi-region deployments using Kubernetes, Terraform, and GitHub Actions.
  • Optimized high-traffic databases (PostgreSQL, ClickHouse) for low-latency retrieval and real-time processing.
  • Maintain a 100% Job Success Score and 'Top Rated Plus' status for consistent high-quality delivery.

What that looks like in practice

At AppsCode Inc. the recurring theme is the number of clusters. A feature that is trivial in one cluster (granting a developer access, collecting metrics, installing an add-on) becomes a distributed systems problem once there are dozens of them across different clouds, some of them private and unreachable from outside. Most of my work has been in that layer: operators that reconcile intent across clusters, APIs that issue scoped credentials, and a monitoring pipeline that keeps each tenant's data separate.

The client work has been the opposite shape: smaller systems, but end to end. I own the architecture, the code, the deployment and whatever breaks at 2am. The clearest examples are the real-time AI systems: a voice agent where speech becomes text, retrieval and a model produce an answer, and the answer becomes audio again inside a window short enough to feel like a conversation, plus the GPU serving layer, autoscaler and video avatar around it.

The same instinct shows up in the smaller things I've built for myself: a Go marketplace service exists mostly because I wanted to work through the asynchronous deployment and billing model properly.

Case study: multi-cluster Kubernetes platform

Case study: real-time voice agent platform

Education

B.Sc. in Computer Science & Engineering

2019 – 2023

East West University

  • Specialized in Distributed Systems and Cloud Infrastructure [CGPA: 3.49/4.0].
  • Thesis: 'Data Security in Cloud Computing using a Hybrid Algorithm Approach'.
  • Published in IEEE Xplore: read the paper.

Related