Available for AI & Backend Engineering Roles

Abdul Khaliq

Backend Systems & AI Engineer

Building end-to-end AI applications, scalable RAG architectures, real-time data pipelines, and production APIs.

Explore Projects GitHub LinkedIn Contact Me

Engineering Works

System Architectures & AI Products

Engineered with a focus on high throughput, zero unnecessary API overhead, sub-second latency, and explainable models.

SaaS & Computer Vision

Iconspire — AI Icon Generator & Background Remover

Zero-cost local segmentation pipeline with end-to-end payment billing workflows

100% Local model execution (ONNX Runtime)

Architecture Highlights

▹ Built local background removal pipelines using Node.js/Express and ISNet ONNX models for image segmentation with zero third-party API costs.
▹ Integrated payment webhooks and sandbox checkouts via Paddle and Lemon Squeezy to handle subscription workflows.
▹ Managed asynchronous image buffers and optimized asset delivery pipelines.

Node.js Express ONNX Runtime Paddle API Lemon Squeezy ISNet Model

Cloud VM & Real-Time

CommuteClear — Real-Time Traffic Alert Pipeline

Automated cloud scrapers, commute window filters, and geofenced event dedup

Architecture Highlights

▹ Configured automated scrapers running on an OCI virtual machine to aggregate live traffic incidents.
▹ Implemented incident deduplication algorithms, commute window alert filtering, and live routing integration.
▹ Designed database schemas and real-time state synchronization using Supabase (PostgreSQL).

Python FastAPI Supabase OCI A1 VM Scrapers PostgreSQL

RAG & LLaMA 3

Sigma RAG — Video Q&A AI Teaching Assistant

Sub-2-second context retrieval with MM:SS video timestamp precision

<2s Average RAG query latency

Architecture Highlights

▹ Built video transcript ingestion and vector embeddings for context-aware Q&A with sub-2-second retrieval speed.
▹ Mapped generated responses directly to exact video timestamps (MM:SS) for precise student navigation.
▹ Integrated Groq API LLaMA 3 acceleration with FAISS vector stores for ultra-low latency inference.

Python LangChain FAISS Groq API (LLaMA 3) FastAPI

NLP & Matching

AI Resume Screening & Candidate Ranker

87% screening accuracy with bias detection filters and semantic parsing

87% Resume & candidate matching accuracy

Architecture Highlights

▹ Built NLP matching and skill extraction algorithms achieving 87% candidate screening accuracy, reducing evaluation time by 75%.
▹ Implemented candidate ranking mechanisms, skill extraction, and bias detection filters.
▹ Structured automated document vector search for multi-format resumes with interactive candidate comparison.

Python NLP Scikit-Learn Streamlit Vector Search

ML & Explainability

Student Performance Predictor & Explainability

84% pass/fail risk classification with deep SHAP feature-importance insights

84% Pass/fail risk classification accuracy

Architecture Highlights

▹ Trained classification models on historical academic data to predict pass/fail risk with 84% accuracy.
▹ Integrated SHAP explainability visualizations to provide transparent feature-importance insights.
▹ Engineered proactive academic intervention triggers and deployed live on Streamlit Cloud.

Python Machine Learning XGBoost SHAP Pandas

Competencies

Technical Skills Matrix

A comprehensive overview of frameworks, infrastructure, and toolsets utilized across real-world systems.

AI & RAG

Generative AI, Vector Retrieval, Local Inference

LangChain

RAG Architecture

Vector Search (FAISS)

ChromaDB

Groq API (LLaMA 3)

ONNX Runtime

Dense & Sparse Embeddings

Prompt Engineering

Backend & APIs

High Throughput Web Services, Asynchronous Tasks

FastAPI

Flask

Node.js

Express

RESTful APIs

Webhooks & Payments

Automated Data Scrapers

Async Buffers & Streams

Databases & Cloud

Distributed Storage, Cloud VMs, Deployments

Supabase (PostgreSQL)

Oracle Cloud (OCI A1 VM)

Docker & Containers

Vercel

Railway

Relational Schema Design

Real-time Sync & RLS

Data Science & ML

Model Training, Feature Attribution, Evaluation

Python

Pandas

NumPy

Scikit-Learn

XGBoost

SHAP (Model Explainability)

NLP & Normalization

Statistical Evaluation

⚡

Low-Latency RAG Ingestion

Chunking semantic hierarchies with sub-2s Groq LPU inference, ensuring immediate retrieval and precise timestamp references.

🛠️

Zero-Cloud-Cost Local Inference

Deploying optimized ISNet and transformer models locally via ONNX Runtime to eliminate prohibitive cloud API inference fees.

🔄

Resilient Real-Time Scrapers

Running headless scraping cron workers on OCI ARM instances with smart deduplication and Supabase PostgreSQL sync.

Direct Communication

Let's Build Something Reliable

Have a project, role, or architecture challenge? Send a message directly or connect across professional channels.

Get in Touch

I am active on GitHub and LinkedIn, and always open to discussing AI systems architecture, contract engagements, or engineering roles.

Email Address engrrabdulkhaliq@gmail.com

Phone & WhatsApp +92 346 8314482

LinkedIn Profile in/engrrabdulkhaliq001

GitHub Repositories github.com/engrrabdulkhaliq001

Response Time: Typically within 24 hours

Send a Direct Message

Have a project, role, or architecture challenge? Leave a message and I'll respond promptly.