identity: Muhammad Maaz Korejo
current_roles:
- "Back-End AI Engineering Intern @ FlyRank.ai (Casper, WY · Remote)"
- "Research / Implementation Engineer Fellow @ INFERENCE Lab (Cohort 01)"
- "AI Developer Intern @ Nebulark (IDP SaaS)"
core_philosophy: "Engineering discipline over hype. Real system output over slide decks."
education: "BS Information Technology, University of Sindh, Jamshoro (2024–2027)"
verified_credentials: "12 Anthropic | 33 Total Accredited Certifications"I am an AI Developer & Backend Systems Engineer specializing in taking LLM-integrated platforms, intelligent document extraction pipelines, and agentic workflows from research papers into production-hardened, low-latency codebases.
- ⚡ Intelligent Document Processing (IDP): Architected hybrid 3-tier routing pipelines that slash cloud API extraction expenses by 80–90%.
- 🔬 Offline LLM Evaluation: Engineered
llm-eval-kiton PyPI, offering zero-cost, deterministic semantic and quantitative evaluations without LLM-as-a-judge dependencies. - 🛠️ Developer Tooling: Authored
ctx-bridgeon PyPI, an open-source context packaging CLI with automated language detection and secret token masking.
| Metric | Achievement | What It Translates To |
|---|---|---|
| 📉 80–90% | Cloud OCR Cost Cut | Tiered fallback: 0.5s LiteParse → CPU Docling → Mistral/Chandra cloud |
| ⚡ 3.3x | Parallel Batch Speedup | Implemented distributed Ray parallel actors for multi-page batch PDFs |
| ⏱️ 292+ hrs | Verified AI Engineering | Completed Back-End AI track + accepted capstone at FlyRank.ai |
| 🧪 100% | Pytest Coverage | Zero-compromise test-driven development across CLI tools & microservices |
| 🛡️ 12 | Anthropic Verified Credentials | MCP, MCP Advanced, Claude API, Vertex AI, Agent Skills, Subagents, Claude Code |
| 📜 33 | Total Verified Certifications | Anthropic, Google, The Linux Foundation, IBM, HP, Forage, UniAthena, PITP |
|
|
|
|
|
|
┌─────────────────────────────────────────────────────────────────────────────────────────────┐
│ EXPERIENCE TIMELINE │
├────────────────────────────────┬───────────────────────────┬────────────────────────────────┤
│ ROLE │ ORGANIZATION │ PERIOD │
├────────────────────────────────┼───────────────────────────┼────────────────────────────────┤
│ Back-End AI Engineering Intern │ FlyRank.ai (Casper, WY) │ Jul 2026 – Sep 2026 (Completed)│
│ AI Developer Intern │ Nebulark (IDP SaaS) │ Jul 2026 – Present (Current) │
│ Research Implementation Fellow │ INFERENCE Lab (Cohort 01) │ Jul 2026 – Present (Fellow) │
│ Cloud Computing Intern │ PITP–MUET (Govt of Sindh) │ Mar 2026 – May 2026 │
│ Certified Graphic Designer │ PITP–MUET (Govt of Sindh) │ Jan 2025 – Mar 2025 │
└────────────────────────────────┴───────────────────────────┴────────────────────────────────┘
🚀 1. Back-End AI Engineering Intern — FlyRank.ai (Casper, WY · Remote | Jul – Sep 2026)
Status:
Completed Track| Verified ID:FR-D11-FA8DB-52256
- Completed 292+ documented engineering hours covering API contracts, task design & prompting, Retrieval & Grounding (RAG), and evaluation operations.
- Culminated program with accepted capstone project "Your 10x Solution", reviewed and accepted by lead mentors and executive leadership.
- Shipped 5 core production assignments: web scrapers, JWT/auth protection layers, multi-container Docker stacks, PostgreSQL database integrations, and CRUD REST APIs.
- Attended 29 live technical masterclasses and completed 55 industry learning resources alongside completing 12 Anthropic Academy certifications.
- Accredited Proof: Certificate of Completion, Recommendation Letter by CEO Alen Malkoc, and Final Evaluation Report by Director Arijana Ibrović.
⚡ 2. AI Developer Intern — Nebulark (IDP SaaS) (Jul 2026 – Present)
Status:
Current Role| Domain: Intelligent Document Processing & OCR
- Executed a rigorous 14-tool benchmark across OCR and layout parsing engines (Docling, Mistral-OCR-4, Chandra-OCR, DocLayout-YOLO, Reducto, LandingAI DPT-3, PaddleOCR, LiteParse).
- Built a 3-tier cost-optimized hybrid extraction pipeline (0.5s LiteParse heuristic check → local CPU Docling → Mistral-OCR-4 / Chandra-OCR fallback), reducing cloud API expenses by 80–90%.
- Deployed Ray parallel actor processing for 3.3x speedup on multi-page PDF documents; fine-tuned DocLayout-YOLO (9 document layout classes) using Label-Studio annotations.
- Conducted OpenMAIC provider-registry architectural review and enforced PostgreSQL Row-Level Security (RLS) tenant isolation.
🔬 3. Research / Implementation Engineer — INFERENCE Lab Fellowship (Cohort 01 | Jul 2026 – Present)
Status:
Cohort 01 Fellow| Project:llm-eval-kiton PyPI
- Selected from 130+ applicants to build
llm-eval-kit, an open-source, pip-installable Python library for offline, deterministic LLM evaluation. - Implemented 4 evaluation modules (Factual Grounding, Relevance, Refusal Detection, Completeness) with zero external judge API calls using local sentence-transformers (
all-MiniLM-L6-v2). - Designed thread-safe singleton model caching and strict numeric-mismatch verification layers to prevent false-positive matches in quantitative scoring.
☁️ 4. Cloud Computing Intern — PITP–MUET, Jamshoro (Mar – May 2026)
- Delivered capstone Flask + PostgreSQL + Railway web platform with automated CI/CD under a government-backed technical training program.
- Decreased lab setup time by ~30% via automated VM configuration and onboarding scripts.
🎨 5. Certified Graphic Designer Intern — PITP–MUET (Jan – Mar 2025)
- Produced 10+ branded visual assets for institutional print and digital communications under an accredited government design curriculum.
- Completed with official Certified Graphic Designer credential.
| Project | Tech Stack | Highlights | Links |
|---|---|---|---|
| 📦 CTX-Bridge | Python Typer Rich pytest PyPI |
Open-source CLI to package codebase context for LLM handoffs (Claude, Cursor, ChatGPT). Secret masking + Claude Code presets. |   |
| 🔬 llm-eval-kit | Python sentence-transformers PyTorch pytest |
Deterministic offline LLM evaluation without LLM-as-a-judge API costs. Grounding, relevance, refusal, completeness. | |
| 🧠 Self-RAG Engine | Python Groq LLaMA 3.3 70B LangGraph Docker |
Self-Reflective RAG (Asai et al.) with 4 reflection checkpoints (Retrieve, IsRel, IsSup, IsUse), query rewrite fallback, 100% pytest coverage. |
|
| 🎙️ Alfred Assistant | Flask LangGraph pgvector Piper TTS React 18 |
Agentic assistant with 14 tools, persistent vector memory, WebSocket live trace visualizer, and local FFT clap-to-wake detector. |  |
| ⚽ AI Pundit Rankings | Flask Supabase PostgreSQL Mistral AI APScheduler |
Social FIFA World Cup 2026 prediction platform with automated AI background grading & community roasting pipeline. | |
| 🛡️ Network IDS | Python XGBoost Random Forest SHAP Scikit-learn |
End-to-end ML pipeline classifying 14 attack categories on CICIDS-2017 with ~99% accuracy and SHAP feature explainability. | |
- Models & APIs: Claude / Anthropic API · Groq API (LLaMA 3.3 70B) · Mistral AI · sentence-transformers (
all-MiniLM-L6-v2) - Agentic Frameworks: Model Context Protocol (MCP) · Claude Code (Subagents & Agent Skills) · LangGraph · DSPy Prompt Optimization
- RAG & Memory: Retrieval & Grounding (RAG) · Self-RAG Reflection Checkpoints · ChromaDB · pgvector · Supabase Vector Memory
- Evaluation:
llm-eval-kit(Deterministic Offline Evaluation) · Factual Grounding · Question Relevance · Refusal Detection
- Engines & Parsers: Docling · Mistral-OCR-4 · Chandra-OCR · LiteParse · MinerU / Unstructured · PaddleOCR
- Vision & Layout: DocLayout-YOLO (Fine-tuned on 9 document classes) · Label-Studio Annotation
- Distributed Computing: Ray Parallel Actor Processing (3.3x speedup on multi-page PDFs) · Multi-tier Hybrid Routing
- Languages & Frameworks: Python (Flask, Flask-SocketIO) · Node.js & JavaScript · Typer CLI & Rich Terminal
- Databases & ORMs: PostgreSQL (Row-Level Security Tenant Isolation) · Supabase · Redis / In-Memory Caching
- API & Architecture: REST APIs · Swagger / OpenAPI Contracts · Webhooks · The Twelve-Factor App · Asynchronous Workers
- DevOps & Cloud: Docker · Railway · Vercel · Git / GitHub Actions · Automated CI/CD Pipelines
- Unit & Integration Testing:
pytestwith 100% test coverage discipline · Mock fixtures & state transition tests - Deterministic Validation: Numeric-mismatch verification layers to catch LLM hallucinated figures
🟧 Anthropic — 12 Verified Credentials (Click to expand)
| Credential Name | Issuing Body | Date Issued |
|---|---|---|
| Model Context Protocol (MCP) | Anthropic | Apr 2026 |
| Model Context Protocol: Advanced Topics | Anthropic | Apr 2026 |
| Building with the Claude API | Anthropic | Apr 2026 |
| Claude with Google Cloud's Vertex AI | Anthropic | Apr 2026 |
| Introduction to Agent Skills | Anthropic | Apr 2026 |
| Introduction to Subagents | Anthropic | Apr 2026 |
| Claude Code in Action | Anthropic | Apr 2026 |
| Claude Code 101 | Anthropic | Apr 2026 |
| Claude Platform 101 | Anthropic | Apr 2026 |
| AI Fluency: Framework & Foundations | Anthropic | Apr 2026 |
| Claude 101 | Anthropic | Apr 2026 |
| AI Fluency for Students | Anthropic | Apr 2026 |
🟦 Google & The Linux Foundation (Click to expand)
| Credential Name | Issuing Body | Date Issued |
|---|---|---|
| Google AI Essentials (5-Course Specialization) | Google · Coursera | Jul 2026 |
| AI Fundamentals | Google · Coursera | Aug 2026 |
| AI for Research and Insights | Google · Coursera | Aug 2026 |
| Inclusive Open Source Community Orientation (LFC102) | The Linux Foundation | Aug 2026 |
🏢 Industry Programs, Honors & Fellowships (Click to expand)
| Credential / Honor | Issuing Organization | Verification Ref |
|---|---|---|
| Backend AI Engineering Internship — Certificate of Completion | FlyRank.ai | FR-D11-FA8DB-52256 |
| Official Recommendation Letter | FlyRank.ai (Alen Malkoc, CEO) | FR-D10-4A2D2-7F7FF |
| Final Internship Report & Evaluation | FlyRank.ai (Arijana Ibrović, Director) | FR-D8-2C53C-C9AA4 |
| INFERENCE Lab Fellowship — Cohort 01 Fellow | INFERENCE Lab | Fellowship Cohort 01 |
| Top Participant — ACT AI SkillBridge | HEC / NAVTTC (Govt of Pakistan) | 2026 |
| McKinsey Forward Program | McKinsey.org | Jun 2026 |
| Certified Graphic Designer | PITP–MUET · Govt of Sindh | 2025 |
🛡️ Cybersecurity & Professional Simulations (Click to expand)
| Credential Name | Issuing Organization | Date Issued |
|---|---|---|
| Mastercard Cybersecurity Job Simulation | Mastercard · Forage | 2026 |
| Datacom Cyber Security Operations Job Simulation | Datacom · Forage | 2026 |
| Your Future in Cybersecurity: The Job Landscape | IBM SkillsBuild | 2026 |
| Getting Started with Cybersecurity | IBM SkillsBuild | 2026 |
| Protecting a Device with Malwarebytes | IBM SkillsBuild | 2026 |
| Cybersecurity and Data | IBM SkillsBuild | 2026 |
| Introduction to Cyber Security | Simplilearn SkillUp | 2026 |
| AI for Beginners | HP LIFE · HP Foundation | 2026 |
🐍 Python Specialization — UniAthena & CIQ (Click to expand)
| Credential Name | Issuer | Date Issued |
|---|---|---|
| Python for Data Science & Machine Learning | UniAthena · CIQ | 2025 |
| Data Structure in Python | UniAthena · CIQ | 2025 |
| OOPs Concept in Python | UniAthena · CIQ | 2025 |
| Functions in Python | UniAthena · CIQ | 2025 |
| Modules, Packages & Files in Python | UniAthena · CIQ | 2025 |
| Data Types & Statements in Python | UniAthena · CIQ | 2025 |
| Basics of Python Programming | UniAthena · CIQ | 2025 |
I am currently exploring AI Developer and Backend Systems Engineering roles.
Whether you're building high-throughput agent systems, cost-optimized IDP architectures, or scalable backend infrastructure, feel free to reach out!
Engineered with precision and discipline. © 2026 Muhammad Maaz Korejo.