Spent quite a few years shipping production systems, from desktop and mobile infra to type-safe research pipelines. However, I'm a builder/integrator at heart and love to experiment, be it ML models or multi-agent coding frameworks as of recent. Most of my work has been centered around customer needs, while some have been purely for me and my people to scratch that itch. You know, just your classic kitchen-sink engineer, always there to solve problems :)
To get a feel of how I work, try the one-line setup of ControlKeel to enforce policies and gates on your raw agent intent. Mantra: AI is a great tool, but I have more faith in the human spirit and the friction we create. My recent work and publications include:
- HERA: Hierarchical Ensemble Reasoning Agent. Google ADK multi-agent system with ensemble voting, Code World Models, and GEPA prompt optimizer
- ControlKeel: Control plane for AI coding agents. Deterministic policy validation and review gates across hosts to catch governance drift with learned observability
- Society of Kawaii: Multi-Agent Debate Arena with Prediction Markets
- Agile LLM Evaluation (IEEE DataPort): A Comparative Analysis of LLMs for User Story Generation
- Microplastic-PFAS Interactions (Hinkley Center): Ensemble QSPR models predicting PFAS-MP sorption; R² = 0.89 on 3,816 compounds
During June 2025, I challenged myself to build one idea a day, which landed a gig to build CVector Energy's agent orchestrator module. Available immediately for full-time roles.
Building Impact
I take research off the page and put it into production. My work ships as reliable, tested software that people use everyday.
I pick projects that solve real problems and I'm looking for a team where I can do more of that.
New tools and frameworks come out weekly. I learn what matters, and build what I enjoy.
Contact
Looking for a software engineer, need consulting on AI/ML, or want to talk research? Reach out.
Follow me on X at @arya_minus, schedule a meeting via cal.com/aryaminus, or send me an email.