About me

I'm an AI-Native Architect and Principal Engineer with two decades of work spanning enterprise AI platforms and international human rights research. My practice centers on building governed, production-grade AI systems - designing the architectures, standards, and tooling that turn AI capabilities into reliable, auditable products.

At a leading global entertainment company I lead enterprise AI platform adoption - architecting governed RAG and semantic search systems, agentic workloads, multi-model gateways, and multi-year technology roadmaps from large use-case inventories.

I founded Sadhira AI & Analytics LLC to build mycontext-ai - a production-grade, MIT-licensed Python library for context engineering with LLMs. The library ships 88 cognitive patterns grounded in 150+ peer-reviewed papers, the Context Amplification Index (an open-source metric that measures LLM output quality improvement vs. a raw prompt), and Requirements-as-Code - an AI-native specification paradigm that turns plain-English intent into machine-checkable product specs and CI-ready trace reports. USPTO provisional patent filed February 2026.

Before AI, I led research, editorial, and advocacy practice across South Asia for organizations including the UN Development Programme, the Open Society Foundations, Human Rights Watch, and the International Center for Transitional Justice.

Current Work

  • Senior GenAI Analyst

    The Walt Disney Company - Disney Experiences
    Dec 2024 - Present

    Enterprise AI platform architecture - governed RAG, agentic workloads, multi-model gateway (Claude, GPT-4o, Gemini), AI workspace with SSO/RBAC, spec-driven development. GitLab CI + Terraform IaC.

  • Founder & Principal Engineer - mycontext-ai

    Sadhira AI & Analytics LLC
    Oct 2024 - Present

    Open-source Python library (PyPI - v0.14.0 - 21 releases): 88 cognitive patterns, 13 LLM export formats, Context Amplification Index, Requirements-as-Code. USPTO provisional patent filed February 2026.

    mycontext-docs.pages.dev

  • USPTO Provisional Patent

    Sole Inventor - Pro se via Patent Center
    February 11, 2026

    "System and Method for Modular Context Assembly and Quality-Optimized Prompt Generation for Large Language Models" - establishes 12-month USPTO priority date for the core innovations of mycontext-ai.

What I Do

  • Enterprise AI Platform Architecture

    Multi-year AI adoption roadmaps, governed agentic and retrieval workloads, architecture decision records, and standards for brownfield and greenfield enterprise systems.

  • Context Engineering & Open Source

    Production-grade context engineering with LLMs via mycontext-ai - cognitive patterns, structured context assembly, CAI measurement, and cross-provider portability.

  • Spec-Driven / Requirements-as-Code

    AI-native specification authoring that turns plain-English intent into machine-checkable product specs, technical specs, and CI-ready trace reports for engineering and QA teams.

  • Governed RAG & AI Gateways

    Enterprise semantic/document search with grounded RAG, multi-provider AI gateways with virtual keys, spend controls, PII handling, guardrails, and observability callbacks.

What Colleagues & Experts Say

  • Dhiraj combines deep technical range, real intellectual curiosity, and the ability to turn ideas into working, well-documented systems. He does not treat architecture, development, testing, and documentation as separate specialties - he takes ownership of the whole arc. He is the kind of engineer any organization is fortunate to have.

    Justin Dial

    Lead Data Engineer · Global Entertainment Company

  • He was not executing tasks - he was architecting solutions. When the program encountered technically complex integration scenarios, Mr. Pokhrel was consistently the resource we turned to for resolution. What distinguished him from other technical contributors was not only his depth of skill but the independence and judgment with which he operated.

    Sindhu J. Manuvalil

    Software Engineering Director · Top-10 US Commercial Bank

  • The combination in mycontext-ai of a structured abstraction, a measurable evaluation metric, and research-grounded patterns constitutes an original contribution of major significance to the field of applied AI and context engineering. CAI reframes prompt evaluation quantitatively - reducing the question of whether structured context produces measurably better output to a single, reproducible figure, making it falsifiable and bringing a discipline ordinarily confined to research into reach of working engineering teams.

    Dr. Rabindra Lamsal

    Postdoctoral Fellow, GRID Lab | Ph.D. Computer Science, U. of Melbourne | ISPRS Best Young Author 2026 · UNSW Sydney

  • Mr. Pokhrel's insistence that a context be a designed, measurable, and portable artefact directly attacks the central deficiency in applied language-model work. The presence of an explicit, reproducible measure of contextual improvement is the feature that distinguishes this work from the majority of prompt-engineering tooling I am aware of, which offers convenience but no evidence. I am myself evidence of its reach: a single article carried this work from the United States to a computer-science department in Delhi and changed how I frame these problems for my students and research group.

    Dr. Pardeep Singh

    Asst. Professor, CS, University of Delhi | Senior Member, IEEE | Ph.D. JNU · Ram Lal Anand College, University of Delhi

Selected Writing