MA

About

About Maarij Aqeel

Full-Stack AI Engineer · Pakistan · working remotely

Maarij Aqeel

I'm Maarij Aqeel, a Full-Stack AI Engineer based in Pakistan, working remotely with teams worldwide. My work sits on the seam between two things that are usually done by separate people: building the machine learning system, and building the product that has to survive real users hitting it.

That combination is deliberate. A model that scores well in a notebook and an application that stays up under load are different engineering problems, and most AI features die in the gap between them. I take responsibility for both ends: the retrieval pipeline and the deploy, the fine-tune and the rate limit.

Today that means voice AI tutors and receptionists, AI automations like a content engine that cut production time 95%, and full-stack AI products like WordBuddy AI, which I took from web app to Google Play while cutting its conversational-AI costs 80%.

Book a call

What I do

  • Voice AI agentsReal-time voice agents that talk to your customers, learners or candidates, and stay affordable at scale.
  • AI automation & agentsWorkflows and agents that take repetitive work off your team, from research to publishing.
  • AI product buildsFrom idea to a shipped, full-stack AI product: web, mobile and the backend behind it.

Experience

  1. Full-Stack AI Engineer

    MelonLabs Ltd · Contract · Mar 2026 – Present

    Building WordBuddy AI, an English-learning platform for Southeast Asian learners. Took it from web app to a Capacitor + Phaser.js mobile app on Google Play, cut hosting costs 60% and conversational-AI costs 80%, and rebuilt offline-first storage on encrypted SQLite with delta sync.

  2. Technical Team Member — AI/ML

    GDGOC FAST NUCES · Part-time · Sep 2025 – Present

    AI/ML technical team at Google Developer Groups on Campus, FAST NUCES.

  3. AI Automation Engineer

    16x9.ai · Contract · Oct 2025 – Feb 2026

    Built an AI content automation engine of four n8n workflows (research, generation, media processing, publishing) that cut content production time 95%, and a Vapi voice receptionist with Airtable-backed knowledge, appointment booking and call routing.

  4. AI Engineer Intern

    DevelopersHub Corporation · Internship · Jul – Sep 2025

    Developed a personalized AI chatbot with context-aware responses and user-specific conversation flows, and used few-shot prompting to make LLM output more consistent and accurate across task-specific workflows.

  5. Full-Stack Developer Intern

    Nexium · Internship · Jul – Aug 2025

    Architected the Intervue AI backend with dual-role auth, cutting voice costs 80% and interview setup time 30% (Next.js, LiveKit, Deepgram, Cartesia, DeepSeek), and shipped an AI blog summarizer that cut content review time 75%.

Education

BS Artificial Intelligence

FAST-NUCES (National University of Computer and Emerging Sciences) · expected 2027

Focus: Machine learning, deep learning and NLP; data analysis (pandas, NumPy, Matplotlib, Seaborn); classical AI; full-stack web development (Django, FastAPI, Next.js); LLM fine-tuning (LoRA, Hugging Face Transformers, DeepSeek, Ollama); and 8086 assembly (NASM).

Languages: English, Urdu.

Other projects

  • Semantic Image SearchMultimodal text-to-image and image-to-image search using OpenAI CLIP embeddings and Pinecone, under 100 ms retrieval across 50,000+ images, with a Gradio UI.
  • CLI AssistantDeepSeek Coder V2 Lite fine-tuned with LoRA to generate command-line tools and automation scripts; published on Hugging Face.
  • UrduFlowLSTM next-word prediction for Roman Urdu (PyTorch), served with FastAPI, with real-time inline suggestions.
  • VisionSnipAI video clipping and subtitling engine (Whisper, FFmpeg, LLMs) that finds high-potential short clips in long-form video.
  • RoleSync AIRésumé optimizer (Next.js, TipTap) that scores résumés against job descriptions, with LaTeX PDF export and a cover-letter generator.
  • SmartDietPlannerAI diet-planning platform with a chatbot.
  • BlogDigesterBlog summarizer with Urdu translation, built at Nexium.
  • Earlier projectsGameStore, a Django games marketplace, and Flappy Runner, a Pygame game where NEAT agents learn to play.

Tech stack

  • Python
  • TypeScript
  • JavaScript
  • React
  • Next.js
  • Node.js
  • FastAPI
  • Django
  • Tailwind CSS
  • Capacitor
  • PyTorch
  • TensorFlow
  • Hugging Face
  • LangChain
  • Ollama
  • Google Gemini
  • OpenRouter
  • scikit-learn
  • LiveKit
  • Deepgram
  • ElevenLabs
  • n8n
  • Make
  • Zapier
  • PostgreSQL
  • MongoDB
  • Redis
  • Supabase
  • Docker
  • Git

Outside of work

  • Cybersecurity (TryHackMe, CTF challenges)
  • Electronics and embedded systems: circuits, microcontrollers, drones and aeromodelling
  • AI game design
  • Chess

Elsewhere

Maarij Aqeel