Cloud-native systems
Kubernetes, admission control, observability, and GitOps work where reliability is designed, measured, and improved under real constraints.
Personal field notes
Cloud-native systems, AI infrastructure, and notes from the build.
I am Abhigyan Mohanta, a software engineer and Open Source Contributor specializing in Kubernetes, AI/ML systems, and the engineering work that turns ideas into runnable, explainable systems.
Open Source Contributor Active contributor to Kubernetes SIG
Scheduling.
AI-native systems Infrastructure, Scaling, Scheduling,
evaluation, and demos.
Learning in public Research notes, build logs, and honest
tradeoffs.
- RL-driven Kubernetes Scheduling
- Real-time Media Pipelines
- Multi-LLM Architectures
- MCP Clients & Voice Agents
- Go & Elixir Backends
- FullStack AI-Native Applications
- Concurrency
01 Core Expertise
Systems engineering with product momentum.
I like work where the fullstack, platform, and AI pieces have to meet reality: latency, failure, cost, team velocity, and users who need the thing to actually run.
AI-native product engineering
Multi-LLM workflows, voice agents, MCP clients, and evaluation-minded prototypes that move from research into usable systems.
Real-time fullstack depth
WebSockets, media streaming, PubSub flows, and front-to-back state synchronization for products where latency and UX both matter.
Builder energy
I like ambiguous problems, technical research, hands-on implementation, and learning in public until the system actually works.
02 Industry Experience
From AI collaboration rooms to social product infrastructure.
Arthur Digital
Added multi-LLM support to Arthur Vibe and developed a Slack/Jira MCP client for automated task updates. Integrated LLMs to build voice agent bots and worked on a DataSync service for keeping virtual room data up-to-date across distributed systems. The product I contributed to was featured on Fox News.
Super Effective
Developed the Select Threader Chrome Extension to bypass expensive API dependencies, saving $5,000 monthly. Set up CI/CD pipelines for biweekly releases and integrated Sentry. Created an Elixir OAuth V2 service, migrated PostgreSQL models, and integrated thread pages into the Next.js web client.
03 Research & Open Source
Academic research and core ecosystem contributions.
Accepted at IC-SIT 2026 (IEEE)
AdmiRL: Learning-Augmented Admission Control for Gang and Elastic GPU Workloads in Kueue
As AI models scale, infrastructure becomes the bottleneck. This paper proposes a reinforcement learning approach for admission control in Kubernetes using Kueue to improve GPU utilization and reduce wait times for large-scale AI workloads. Co-authored with Saswat Sankalp Parida, Rakesh Kumar Parida, and Smruti Sikha Dash, under the guidance of Dr. Barnali Sahu.
Open Source Contributor
Kubernetes SIG Scheduling & Kueue
Active contributor to Kubernetes SIG Scheduling, focusing on the Kueue project and KueueViz. Recent contributions include mitigating DoS vulnerabilities via HTTP and WebSocket safeguards, fortifying secure deployments, resolving dashboard crashes, and enforcing container security best practices.
04 Selected projects
Open-source tools, interactive systems, and side projects.
Open Source
K8s GitOps Delivery Platform
A comprehensive Kubernetes GitOps delivery platform featuring automated canary releases with Argo Rollouts, event-driven autoscaling with KEDA, and CI/CD pipelines via Jenkins.
Game
Escape the Fall Season
A fun 2D arcade game built in C++ and SFML. Play as a character avoiding falling logs and branches with dynamic spawning and intuitive controls.
Archived
UnbreakableMe
A motivational Android app inspired by David Goggins. Designed to help tap into inner resilience with daily quotes, goal tracking, and mindfulness exercises.
05 Recommendations
What managers say about working with me.
These are pulled from recommendations on my LinkedIn profile, with each recommender linked directly.
View all LinkedIn recommendationsDinesh babu
Abhigyan is a proactive problem solver and the go-to person for tackling complex challenges that demand technical research and hands-on implementation.
Joze Kosmerl
He is a great engineer. Very resourceful, hard working, and smart guy. He would be a great addition to any engineering team.
06 Let's talk
Have a system to build, scale, or debug?
Book a 30-minute call if you want to talk through cloud-native infrastructure, AI product engineering, and architecture. I also love geeking out on industry trends and best practices. Feel free to reach out!