I'm a Data Scientist and Associate Software Engineer focused on architecting intelligent, production-grade systems — I care more about pipelines that ship than ones that only work in a notebook. At Acme One, I contribute individually to Nucleus One, the company's core multi-module enterprise platform, after building my foundation by shadowing senior engineers across other internal systems. Independently, I'm building agentic AI workflows with LangChain and LangGraph — multi-step chains, tool-calling agents, and stateful pipelines that ship to production — including a Retrieval-Augmented Generation platform aimed at solving document-heavy bottlenecks for professional services and local firms. I'm currently deepening that expertise toward the Microsoft AI-103 (Azure AI Apps & Agents) certification.

⭐ If any of these save you time or teach you something, a star helps more than you'd think — it's what makes them discoverable to the next person who needs them.

RepoWarden — Agentic AI Code Engine

Points a two-agent LangGraph system at any local codebase: a read-only Reviewer proposes changes, and a separate write-capable Editor is only ever constructed after you explicitly approve — the safety boundary is structural, not a prompt instruction the model has to keep honoring.

• Two compiled LangGraph StateGraphs (Reviewer, Editor) sharing MongoDB-backed short- and long-term memory
• Token-level streaming over Server-Sent Events, with structured tool-call/analysis/edit events
• FastAPI backend · React + Monaco-editor frontend · custom tool definitions

Repo: RepoWarden

MFA Authentication Server — Information Security capstone

Want to actually understand how TOTP/HOTP work instead of just importing a library that does it for you? Every OTP algorithm here is implemented straight from the RFC spec, by hand.

• HOTP (RFC 4226) and TOTP (RFC 6238) implemented from the spec — validated against pyotp as an independent test oracle, never used in production code
• AES-256-GCM encrypted secret vault, Argon2id password hashing, real WebAuthn/FIDO2 attestation and assertion verification
• AI anomaly-detection layer — IsolationForest scoring over login features (geo distance, impossible travel, new device, unusual hour) with Claude-generated plain-language alerts
• Streamlit admin dashboard for users, auth logs, and anomaly alerts

Repo: Multi-Factor-Authentication-Protocols

Document Intelligence Platform for Professional Services (In Development)

Ask your firm's own documents a question in plain English and get a cited answer back — instead of digging through folders. Built for law firms and accounting practices buried in paperwork.

• Retrieval — Azure AI Search over indexed firm documents
• Generation — Azure OpenAI Service, grounded strictly in retrieved context
• Ingestion — Azure AI Document Intelligence + Blob Storage for multi-format document processing
• Trust layer — inline citations back to source paragraphs on every answer
• Access control — Microsoft Entra identity integration

Built on top of Microsoft's production RAG reference architecture, which I forked and am extending toward this vertical use case. Repo: Project-Ease

Automated Information Security Risk Assessment Platform — UMT InfoSec semester project, Spring 2026

Turns a NIST SP 800-30 risk assessment from a spreadsheet exercise into a working app — asset inventory, live CVE matching, and auto-generated reports. Handy if you're studying for a security certification or building your first GRC tool.

• Risk engine covering asset, threat, and vulnerability management with live CVE lookups via the NVD API
• LLM-generated control recommendations (Claude, with a rule-based fallback when no API key is configured)
• Automated PDF reporting — risk register, cost-benefit analysis, and compliance checklists
• Flask REST API backing a Streamlit UI, with a React frontend also included

Repo: Automated-Information-Security-Risk-Assessment-Platform

Acme One · Lahore, Pakistan · 1 yr+

Associate Software Engineer · Full-time · Mar 2026 – Present Contributing independently on Nucleus One — a production enterprise platform built on a React 18/TypeScript frontend and ASP.NET Core 8 / SQL Server backend — as one of the most active engineers on a 7-person team, #2 by commit volume across both repositories.

• End-to-end delivery: built Project-One entirely — database schema, backend logic, and frontend, including core workflow components like the work item management drawer
• Service architecture: authored core HR module services — employee onboarding, offboarding, and evaluation workflows — plus its database schema deployment
• Platform security: built the authentication controller for the Nucleus One core platform
• Production deployment: took TimeTrace to production, including its backend reporting engine

Artificial Intelligence Intern · Jul 2025 – Mar 2026 Engineered and deployed a Retrieval-Augmented Generation (RAG) chatbot directly into Nucleus One, enabling natural-language interaction with internal company data — the internship converted into the full-time Associate Software Engineer role above.

• AI-103 — Designing and Implementing a Microsoft Azure AI Solution (in progress)
• Claude Code in Action — Anthropic, 2026
• Claude Code 101 — Anthropic, 2026
• AI Fluency Framework & Foundations — Anthropic, 2026
• AI Capabilities and Limitations — Anthropic, 2026
• Claude 101 — Anthropic, 2026

_{3D isometric render of my commit history}

_{A jet flying over my real contribution grid, firing on the busiest days}

Discipline outside the IDE is the same discipline that ships code on time.

"Sab bikta hai, sab khareedoon ga." — everything is for sale, and I'll buy it all.