I design human-reviewed AI quality systems that improve engineering speed while preserving traceability, deterministic controls and accountable release decisions.
My work combines Quality Engineering architecture, prompt and context engineering, agent workflow validation, CI/CD quality gates and technical program leadership.
Prompt & Context Regression · Source Grounding · Deterministic Validation · Tool-Use Testing · Human Approval
Current official designation: Technical Project Manager
AI can accelerate engineering work, but speed alone does not produce trustworthy outcomes. Quality comes from controlling sources, context, tools, validation and approval.
Responses may be plausible but unsupported by approved requirements, source material or system evidence.
Prompts, retrieved content and conversation state can change behavior without a visible code change.
Agents may call the wrong tool, use the wrong sequence, alter unintended data or continue beyond their approved boundary.
Approved source material, explicit context assembly, version tracking and traceable references.
Schemas, assertions, rule checks, regression datasets and repeatable failure analysis.
Explicit human decision points, tool permissions, reviewable logs and release evidence.
A controlled delivery lifecycle from source evidence through release approval.
Requirements, acceptance criteria, risk and success conditions
Approved sources, versioned context and privacy boundaries
AI-assisted content, code, tests, API assets or agent tool execution
Schema checks, deterministic assertions, scenario regression and traceability
Engineering judgment, security review and explicit human approval
Publish approved outputs, retain evidence and use failures to improve the evaluation set
The portfolio distinguishes operational work from product demonstrations and architectural methods.
Used within governed engineering workflows.
Demonstrated through Aarohan CareerOS and selected engineering platforms.
A reusable approach for trustworthy AI and agent delivery.
Four connected disciplines convert unpredictable model behavior into reviewable engineering signals.
Version prompts, system instructions, retrieved context and expected behavior. Re-run representative scenarios when any input changes.
Regression suite · Version comparison · Failure taxonomy
Verify that outputs are supported by approved sources, meet required structure and avoid unsupported claims or missing evidence.
Source traceability · Schema validation · Deterministic assertions
Test tool selection, input parameters, execution order, permission boundaries, retries and recovery from partial failure.
Tool-call traces · Workflow checks · Boundary tests
Combine evaluation scores, deterministic checks, human approvals and failure evidence before releasing an AI-assisted workflow.
Approval record · Quality gate · Audit evidence
Evidence across employment workflows, award-recognized team delivery and personal engineering platforms.
Transforms requirements and screenshots into structured documentation, delivery tasks, manual tests, Postman collections, traceability and quality metrics.
Controls: Human review · Output structure checks · Source traceability · Approval boundaries
A local-first career evidence and application platform built around governed sources, reusable assets, privacy controls and reviewable human decisions.
Evidence: Next.js · FastAPI · SQLAlchemy · API tests · Playwright · GitHub Actions
A Playwright API automation solution with reusable validation, reporting, CI/CD execution and controlled AI-assisted engineering.
Evidence: Semicolons 2026 Spot Award · Team Dhurandhar
Transforms approved source material into age-adapted stories, expressive audio, images and activities with automated validation and controlled publishing.
Evidence: Google and ChatGPT APIs · ElevenLabs · GitHub Actions · Google Drive · WhatsApp delivery
AI quality is not only a model problem. It is an operating-model, people and accountability problem.
Technical leadership, prioritization, review and mentoring
Delivery across the United States, India, the Philippines and Mexico
Automation, API, performance and responsible AI-assisted engineering
Selected AI-assisted activities with mandatory human review and engineering controls
Generated code, tests, documents and API assets remain subject to engineering review
Prompt patterns, contribution standards, onboarding curricula and knowledge systems
I combine Quality Engineering architecture, AI evaluation controls, full-stack technical depth and technical program leadership. The result is a practical approach to AI adoption: measurable acceleration where appropriate, explicit controls where risk matters and human accountability throughout.
Reusable automation, performance, API, mobile, data and CI/CD quality systems
Prompt and context regression, grounding, deterministic validation, tool-use testing and approval gates
Roadmaps, sequencing, risk, executive communication and distributed engineering enablement
Current official designation: Technical Project Manager