AryanKumar
3+ years shipping production AI infrastructure in NSA-compliant, air-gapped environments. Started with full-stack apps. Now I own platform workloads that used to take five engineers.
Stack
Technologies & Skills
Go, Python, TypeScript, Kubernetes, LiteLLM, Helm, RAG, PyTorch, Next.js, FastAPI, Redis, PGVector, Terraform, Grafana.
AI/ML Engineer
Specialized in neural network development, LLM deployment, and AI infrastructure
Core Expertise:
Full Stack Developer
Building scalable web applications with modern frameworks
Languages
TypeScript
Python
Java
Go
Frontend
React
Next.js
Tailwind
DevOps & Infrastructure
Kubernetes
Docker
Terraform
AWS
Backend & Databases
Spring Boot
Flask
Node.js
MySQL
Additional Skills
Currently Exploring
Advanced AI model optimization, distributed systems, and cloud-native architectures
Experience
Career Arc
From Python scripting to platform engineering in two years at the same company.
// before the clusters
Python Developer
Expert Witness Profiler LLC
Jun 2022 – Sep 2022
Built automated data collection pipelines using Beautiful Soup and Selenium for a legal tech client. Developed Pandas-based processing workflows for court records analysis. First production software work — three months, then straight into NALEJ.
// full-stack → infrastructure
Software Engineer
NALEJ.AI
Sep 2022 – 2024
Joined to ship full-stack tooling — built a GitLab analytics portal (Next.js, tRPC, Keycloak, Redis) with full offline compatibility and a deep learning issue estimation system using semantic embeddings. Quickly expanded into infrastructure: architected air-gapped LLM deployments using LiteLLM, Ollama, and Kubernetes, cutting setup time from hours to under one minute. Built Helm automation frameworks and contributed to a custom Kubernetes operator written in Go.
// platform engineering
Platform Engineer & Tech Lead
current
NALEJ.AI
2024 – Present
Took ownership of a critical platform workload previously staffed by five engineers — leading design, incident response, and operational tooling. Owns cluster-wide Karpenter and KEDA configurations for automated node provisioning and event-driven autoscaling. Delivered end-to-end DevOps for a federal defense client across EKS and K3s: provisioning, image lifecycle, and air-gapped package distribution. Platform work contributed to $15.6M/year in operational savings. Ensured NSA-level compliance across all deployments. Currently building an agentic AI system for natural-language Kubernetes automation.
Let's work together
Building something in a constrained environment, or just want to talk shop?