Independent Engineer & Product Builder
AI Systems | AI Evaluation | Healthcare Technology | Agent Reliability | Automation | Verifiable Infrastructure
I build practical software around problems where AI, real-world operations and dependable infrastructure meet.
My current work spans AI and agent systems, evaluation and reliability, healthcare workflows, local and hybrid inference, automation, privacy-aware systems and verifiable digital infrastructure.
- AI Evaluation & Reliability - reproducible evaluations, model and agent testing, evidence-based verification, failure analysis, disagreement resolution and regression testing.
- Agent Safety & Control - deterministic policy enforcement, approval boundaries and human oversight for AI agents.
- Healthcare Technology - hospital operations, claims workflows, documentation, patient continuity and AI-assisted operational systems.
- AI Infrastructure - local-first and hybrid inference, routing, automation and cost-aware execution.
- Verifiable Systems - auditable workflow proofs, privacy-preserving verification and accountable digital records.
Deterministic policy and approval infrastructure for AI agents, separating model interpretation from authority and execution.
https://github.com/kaulastudies/boundary-ai-agent-control
Local-first and hybrid AI routing focused on privacy, reliability, token efficiency and intelligent execution decisions.
https://github.com/kaulastudies/omniedge-routeiq
A structured pre-mortem system for exploring failure paths, assumptions, safeguards and decision readiness before execution.
https://github.com/kaulastudies/failuretwin-ai
Verifiable workflow-proof infrastructure using cryptographic hashing and Stellar/Soroban for accountable real-world events.
https://github.com/kaulastudies/ProofSetu-Protocol
Local AI agent experimentation focused on private execution and explicit permission before sensitive actions.
https://github.com/kaulastudies/boundary-amd-private-agent
A collaborative Reddit-native application built with Devvit, persistent shared state and role-based user experiences.
https://github.com/kaulastudies/thread-zero
Working Alpha / Repository Agent Reliability Program: provider-neutral infrastructure for evaluating AI systems and agents through versioned task contracts, controlled execution, independent verification, verifier testing, canonical evidence records, source-exact replay and regression analysis.
https://github.com/kaulastudies/ai-evaluation-lab
Current public RARB evidence includes 51 promoted live attempts across five verifier-qualified repository tasks, plus a preregistered 30-attempt cross-model replication on one frozen task with source-exact no-model replay. Claims are intentionally bounded to the committed tasks and preserved evidence.
A public-safe engineering dossier for healthcare operations, claims, care continuity and exception-driven hospital workflows developed from real operational problems.
https://github.com/kaulastudies/claimsetu
The public repository intentionally excludes patient data, hospital-confidential material, production credentials and sensitive deployment details.
I contribute targeted fixes, tests and maintenance work to upstream open-source projects in developer infrastructure and AI evaluation.
- Backstage Community Plugins / Periskop - migrated the workspace to Jest 30. Maintainer-approved and merged into
main: backstage/community-plugins#10964
- Kubernetes SIGs Headlamp - stale-response protection and regression coverage for KubeObjectGlance: kubernetes-sigs/headlamp#7659
- MLCommons ModelBench - dynamic SUT listing for OpenAI and Together dedicated: mlcommons/modelbench#1650
- MLCommons ModelBench - Mistral native model-ID formatting fix: mlcommons/modelbench#1661
- MLCommons ModelBench - Anthropic refusal and thinking-readiness handling: mlcommons/modelbench#1662
- OpenFeature Operator - regression coverage for the pod annotation index path: open-feature/open-feature-operator#859
I care about what happens after a prototype works: failure modes, edge cases, security boundaries, evaluation, auditability, maintainability and whether a system can survive real operational use.
My work typically involves TypeScript, Next.js, React, Python, FastAPI, PostgreSQL, Supabase, AI/LLM systems, APIs, automation and Stellar/Soroban.
@kaulastudies is simply the GitHub handle I currently use for technical work.
The software repositories and engineering projects presented here are my independent technical work as Rama Chandra. They should not be interpreted as projects owned by or operated under Kaula Studies.
Kaula Studies is a separate cultural and traditional identity and is kept separate from the software, AI, healthcare and engineering work shown on this profile.
If this work is useful to you, you can support my independent public engineering and applied research through either route:
- GitHub Sponsors: https://github.com/sponsors/kaulastudies
- Direct support via PayPal: https://www.paypal.com/paypalme/malayanur92
Support helps fund public development, evaluation infrastructure, testing, documentation, hosting and maintenance.
Funding does not automatically grant private repository access, contributor permissions, consulting, custom development, confidential data access or exclusive IP rights.
I am interested in serious work around AI evaluation, agent systems, healthcare technology, developer infrastructure, automation and applied AI research.