AI/ML Engineer · Backend Developer · Computer Vision & LLM Systems

Rector's List (3x) · Dean's List (2x) 🏆
I ship production ML and backend systems: from fine-tuned LLM pipelines and computer vision models to FastAPI services with real concurrency and auth concerns.

   

/*
 * ============================================
 *   SYSTEM BOOT :: kinza_afzal.exe
 *   Compiling with -O3 -Wall -std=c++23
 * ============================================
 */

#include <iostream>
#include <vector>
#include <string>
#include <memory>

using namespace std;

class Engineer {
private:
    // Singleton -- there's only one of me, and I ship.
    static inline unique_ptr<Engineer> instance = nullptr;

    Engineer() = default;

public:
    string name      = "Kinza Afzal";
    string role      = "AI/ML Engineer & Backend Developer";
    string location  = "Karachi, Pakistan";
    string education = "BS CS @ FAST NUCES | Rector's List x3, Dean's List x2";

    vector<string> shipped = {
        "[CV]      PPE-compliance pipeline    -> RF-DETR, 97.8% mAP@50",
        "[LLM/RAG] Medi Query                 -> 92% relevancy, 153 tests passing",
        "[ML]      Network Logs Analyzer      -> 96%+ anomaly detection, 494K+ logs",
        "[DL]      FloodDepth Estimation      -> Attention U-Net + ZoeDepth, MLOps end-to-end",
        "[BACKEND] Invite Flow API            -> Node.js/Express, 12+ RLS policies, OAuth",
        "[SYS]     SLA/Penalty Engine         -> FastAPI + PostgreSQL, concurrency-safe"
    };

    vector<string> stack = {
        "Python", "FastAPI", "PostgreSQL", "LangChain",
        "LangGraph", "PyTorch", "TensorFlow", "Docker",
        "Node.js", "Express", "MongoDB", "REST API"
    };

    vector<string> backend_focus = {
        "System Design    -> scalable service boundaries, API contracts, caching layers",
        "Node.js/Express  -> REST APIs, middleware, auth (JWT/OAuth), rate limiting",
        "MongoDB          -> schema design, aggregation pipelines, indexing strategy",
        "Message Queues   -> async job processing, event-driven pipelines, retries/DLQs",
        "Concurrency      -> row-level locking, SLA/penalty engines, race-condition-safe writes",
        "DevOps           -> Docker/Compose, CI/CD (GitHub Actions), AWS EC2, MLflow, monitoring"
    };

    static Engineer& getInstance() {
        if (!instance) instance = unique_ptr<Engineer>(new Engineer());
        return *instance;
    }

    friend ostream& operator<<(ostream& os, const Engineer& e) {
        os << ">> booting " << e.name << " ...\n";
        os << ">> role     : " << e.role << "\n";
        os << ">> location : " << e.location << "\n";
        os << ">> status   : compiling ideas into production systems\n\n";
        os << ">> git log --oneline --author=\"kinza\"\n";
        for (const auto& log : e.shipped)
            os << "   " << log << "\n";
        os << "\n>> cat backend_focus.txt\n";
        for (const auto& item : e.backend_focus)
            os << "   " << item << "\n";
        return os;
    }

    void ship() const {
        cout << *this;
        cout << "\n>> return 0;  // and ready for the next commit\n";
    }
};

int main() {
    Engineer& me = Engineer::getInstance();
    me.ship();
    return 0;
}

/* ============================================
 *   Process finished with exit code 0
 *   Uptime: since 2023 | Status: still compiling
 * ============================================
 */

• AI Product Engineer Intern @ Agento: building on an enterprise AI operations platform (React 18, TypeScript, Vite); shipping features and fixing bugs across the Skills Engine, Model Router, and Agent Shield (RBAC/ABAC policy engine), and partnering with the compliance team on the Evidence Center's audit-trail and governance flows for launch
• AI/ML Engineer Intern @ SYSLAB.AI: computer vision pipelines (OpenCV, Roboflow) for hospital PPE-compliance and anomaly detection; fine-tuned Qwen2.5-3B for automated interview psychometric scoring
• Software Development Engineer (Research) @ FAST NUCES KHI: multi-agent AI replanning system (LangGraph, CrewAI, Neo4j) benchmarked against 7 classical optimization algorithms; scalable FastAPI/PostgreSQL backend with concurrency-safe SLA/penalty engines
• Former AI/ML Trainee @ Shark Stack: AI voice receptionist (Vapi.ai + n8n) with tool-calling workflows; RAG chatbot (LangChain, Pinecone, Groq) at 88%+ response accuracy, Dockerized CI/CD on AWS
• Teaching Assistant @ FAST NUCES KHI: Data Structures & Database Systems; mentored 40+ students, graded 20+ semester projects
• Specialties: computer vision, LLM fine-tuning & RAG pipelines, multi-agent systems, backend architecture with row-level security & concurrency control
• Open to AI/ML Engineering, Backend roles, and research collaborations

• Medi Query: AI Medical Assistant: Medical RAG chatbot (Llama 3.3 70B, hybrid MMR+BM25 retrieval, cross-encoder reranking) with 92% relevancy, 88.7% faithfulness, and 153-test coverage across unit/integration/security/performance, deployed via Dockerized CI/CD on AWS EC2.
• Network Logs Analyzer: Real-time SOC dashboard powered by a GHF-ART clustering model (96%+ detection accuracy across 494K+ network connections), with live threat visualizations, ChatOps analysis, and automated critical-alert escalation.
• FloodDepth Estimation: Flood detection & risk assessment system using Attention U-Net segmentation (86.95% Dice/F1, 76.91% IoU) with ZoeDepth metric depth estimation and Grad-CAM explainability, deployed via a full MLOps pipeline (MLflow, FastAPI, Gradio, Hugging Face Spaces).
• Invite Flow: Supabase Postgres backend with 4 relational tables, 12+ Row-Level Security policies, and security-definer functions; email/password + Google OAuth auth and QR-based invitation/check-in flows, deployed on Vercel with CI-driven multi-environment config.
• PixelLang Compiler: A retro-inspired domain-specific language for pixel art generation, with a complete compiler pipeline (DFA-based lexer, LL(1) recursive-descent parser, 37-rule semantic analyzer, Pillow-based codegen) and a full Tkinter GUI IDE with syntax highlighting and live preview. Built for CS4031 Compiler Construction.
• Gravity Switch: A fast-paced 2D auto-runner mobile game (React Native + Expo, TypeScript, Zustand) where a single tap flips gravity to dodge neon obstacles in a cyberpunk world, with progressive difficulty, particle effects, and haptic feedback, built end-to-end with EAS builds.
• Traffic Simulation System: A multi-threaded traffic simulation (C, POSIX pthreads) modeling traffic flow across multiple intersections, with mutex/condition-variable synchronization, priority scheduling for emergency vehicles, inotify-based hot-reloading config, and a real-time ncurses GUI with live system performance monitoring. Built for an Operating Systems course.

• ImportOptimizer: AST-based npm codemod (TypeScript, Babel scope analysis, magic-string) that detects whole-library imports and safely rewrites them into tree-shakable direct imports, preserving original formatting.

• Data Security Posture Management (DSPM) in the Era of Generative AI and Agentic AI: Toward an Intelligent Agentic DSPM (IADSPM) Framework: Technical Note, ResearchGate. Proposes an intelligent, agentic approach to Data Security Posture Management (DSPM) for securing data across generative and agentic AI systems.

Pair Extraordinaire — @kinza7124 coauthored commits on merged pull requests.