Open to opportunities

Abdullah bin Mansoor

@abdullahbinmansoor

Message

Python Engineer focused on LLM/RAG agents and deployable computer vision solutions.

Pakistan

Message

What I'm looking for

I’m looking for a team where I can build LLM/RAG agent systems and deploy computer-vision models with measurable performance—clean engineering, fast iteration, and ownership of end-to-end pipelines from data to inference.

I’m a Python-focused engineer who builds AI systems that actually ship: agentic LLM pipelines (RAG, Knowledge Graph SQL agents, tool/function calling) and computer vision models engineered for real-world inference. In internships, I delivered a RAG + Knowledge Graph SQL agent that turns natural language into accurate SQL and improved QA automation by revamping Selenium scripts.

My computer vision work centers on clinical-grade pipelines and reliable deployment. I built an ECG digitization pipeline with lead segmentation + waveform extraction reaching 99.4% mAP@50, and an echocardiogram EF estimation system using YOLOv11 keypoint tracking feeding a ResNet-18 regressor (R² ≈ 0.55–0.56), including fixes for early overfitting via composite losses and outlier filtering. Beyond CV, I’ve created multi-agent platforms (e.g., crypto intelligence with ReAct + RAG + MCP tool contracts) and fully local voice AI for streaming ASR/LLM/TTS.

Experience

Work history, roles, and key accomplishments

Python Engineer Intern

Logic Powered Solutions

Sep 2025 - Oct 2025 (1 month)

Built a RAG + knowledge-graph SQL agent to translate natural-language queries into accurate SQL and revamped Selenium automation scripts to reduce manual QA effort. Interviewed 3 engineering candidates and delivered 2 hands-on workshops for a cohort of 5 interns.

RAG Knowledge Graphs SQL Python Selenium SQL Agent Test Automation

AI/ML Intern

AIGenMat

Jun 2025 - Aug 2025 (2 months)

Built and tuned an ML classifier achieving 98% accuracy with a CNN and 85% accuracy with a Decision Tree for materials classification. Automated scraping of 15,000+ business leads using BeautifulSoup and Selenium to significantly reduce manual prospecting time.

CNN Decision Trees Python BeautifulSoup Selenium Web Scraping & Model Tuning Data Pipelines

Computer Vision Intern

DataInsight Research Lab

Jun 2025 - Aug 2025 (2 months)

Developed an ECG digitization pipeline using YOLOv11 lead segmentation and U-Net waveform extraction, achieving 99.4% mAP@50 for class-specific lead detection and 93%+ mAP@50 with ONNX/TFLite edge deployment variants. Built an EchoNet-Dynamic EF estimation pipeline using a YOLOv11 keypoint model and a dual-output ResNet-18 regressor (R²: 0.55 ESV, 0.56 EDV), improving generalization by resolving o

YOLOv11 U Net ResNet 18 ONNX Runtime TFLite NeuroKit2 Object Detection ResNet18

Education

Degrees, certifications, and relevant coursework

FAST-NUCES (Islamabad)

Bachelor of Science (BS), Artificial Intelligence

2023 - 2027

Grade: CGPA: 3.35/4.0

Activities and societies: Dean's List of Honor — 2×

Pursuing a Bachelor of Science in Artificial Intelligence at FAST-NUCES (Islamabad) from 2023 to 2027. Holds a CGPA of 3.35/4.0 and was named to the Dean's List of Honor twice.