Skip to content
View nrupala's full-sized avatar
🎯
Focusing
🎯
Focusing

Block or report nrupala

Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
nrupala/README.md

Nrupal Akolkar

Electrical and Systems engineer building trustworthy, verifiable AI.

P.Eng. with 24+ years in safety-critical energy systems (CNRL, Shell Canada), electrical power generation, distribution and utilization.

Curious and motivated to make a difference and creating value by developing products and services that benefit humanity.

Building open-source tools that bring institutional-grade analytical power to everyone.


What I'm Building

Project What It Is Tech
KalaBodha Personal AI assistant powered by a knowledge graph you own — Dimensional Intelligence Python, Neo4j, Ollama, AES-256-GCM
PAE Personal Analytics Engine - institutional-grade financial analytics for individuals Rust, Python, Plotly, Skfolio
Guardian Mesh AI security governance - 5-enclave deterministic enforcement, formally verified Rust, gRPC, ML-KEM, Lean 4
SAIC Sovereign AI Core - deterministic reasoning that proves its answers Rust, Python, Z3, Neo4j, OWL 2
Research Analyst Automated multi-source research and report generation Python, FastAPI, Capacitor

All projects: AGPL-3.0. Self-hosted. User-owned data. Encrypted with your keys.


Where I'm Headed

Beyond models and agents - toward AI that proves its answers, knows its limits, and earns trust through transparency.

  • Trustworthy. Every answer is traceable. Graph-sourced facts are cited. Model-assisted content is labeled.
  • Verifiable. Open source. Auditable audit trail. The code IS the methodology.
  • Viable. Runs on your hardware. Works offline. No subscription required for full functionality.

Design Principles

  • Tool calculates, user decides. Technology should make people more capable, not more dependent.
  • Privacy by architecture. Your data is encrypted with keys only you hold. This is how the system is built, not a policy that could change.
  • Honesty over confidence. VERIFIED or OUT OF SCOPE. Never a confident guess.
  • Open source is the methodology. The code IS the audit trail.

How to design systems that reach their outcome on purpose, not by luck — a worked theory of convergent transformation, from control loops to financial markets.
Download my book on this topic: Output Convergence System
Relevant blog: Devinfo.dev Mirrored on AIMLDS

Blog on Power System and Smart Grids: Onsmartgrid
Educational Product on market research: Research Analyst
Educational product on Budget Planning and Macro economics: BEE
Encrypted Zero Leakage Notes Taking App: CipherNotes


Background

24+ years of building systems where accuracy and safety are non-negotiable. The discipline that keeps a power grid running is the same discipline that makes an AI system trustworthy.

  • Electrical Engineer (P.Eng.), APEGA Member
  • Canadian Natural Resources (CNRL) - 2017 to present
  • Shell Canada Energy - 2014 to 2017
  • B.Eng. | MBA | Harvard Business School Online | IIM Calcutta
  • Calgary, Alberta, Canada

Connect


Building tools that bring institutional-grade AI to individuals - trustworthy, verifiable, and entirely yours.

Pinned Loading

  1. mykey mykey Public template

    A zero-trust password manager. A NIST/CISA Compliant Canadian Vault v1.0. 🍁 Built in Canada 🍁

    Python

  2. task-habit-tracker task-habit-tracker Public

    Standalone localized task and habit tracker, with encrypted backup and import options.

    HTML 1

  3. vault-tracker vault-tracker Public

    Zero trust self-contained, offline, vault based expense tracker (ledger), habit tracker, notes engine, and tasklist with reminder notification.

    TypeScript 2

  4. axiomcode axiomcode Public

    AxiomCode converts natural language descriptions of algorithms into mathematically proven-correct code in Python and C. Every generated artifact comes with a cryptographic certificate of verificati…

    Python 1

  5. multi-cloud-rag-free multi-cloud-rag-free Public

    Multi-Cloud Free Tier RAG Training Platform Oracle 24GB/400GB + GCP/Azure/AWS micros; K8s-federated Haystack/Milvus/Ollama

    Shell