M.Sc. Applied Computer Science graduate (SRH Heidelberg, 2025). I build the pipelines between systems — computer vision inference models (YOLOv8 + OpenVINO for 25+ FPS on CPU), scalable scraping engines (Selenium, Apify, PostgreSQL), and automations that turn manual busywork into scripts that run themselves.
Raw webcam frames and unstructured dynamic web HTML entering the pipeline.
I graduated with an M.Sc. in Applied Computer Science from SRH Hochschule Heidelberg (Grade 2.2), where my Master's Thesis was an accessible Object Detection Game for Hearing-Impaired Children — utilizing custom YOLOv8 models, quantized via Intel OpenVINO for 25+ FPS on standard CPU laptops without a GPU, and wrapped in a responsive PyQt6 interface.
Since then, I have focused on high-throughput backend services and automations: deploying Python Selenium scrapers as cloud Apify Actors, processing payloads through Make.com JSON pipelines, and syncing relational PostgreSQL databases.
Prior to my Master's, I spent three years in technical support and agile coordination roles — reading server logs over the Linux CLI, running structured SQL queries against production databases, and validating release acceptance criteria. That foundation is why I engineer systems that hold up reliably in production, outside the notebook.
Every technical skill tied directly to a deployed project or production architecture.
Orchestrated 4-agent team (Planner, Hybrid Retriever, Verifier, Synthesizer) with NLI fact-checking.
Engineered AES-256-GCM instant key revocation for GDPR Article 17 Right to Erasure without re-indexing.
Automated benchmark suite measuring Faithfulness (94.5%), Relevance, and Recall under EU AI Act Art. 15.
Trained custom object detection models for real-time visual learning.
Optimized YOLOv8 runtime to achieve smooth 25+ FPS on standard CPU laptops.
Engineered accessible desktop interface with instant large-font visual labels.
Model training, transfer learning, and evaluation for computer vision pipelines.
Automated browser workflows with dynamic pagination and anti-bot mitigation.
Containerized cloud scrapers harvesting 2,000–4,000 records per scheduled run.
Architected 4-app multi-account OAuth sync: Forms ➔ Sheets ➔ Calendar ➔ Gmail.
Built asynchronous REST APIs and backend data processing scripts.
Structured schema design, JSON payload validation, and SQL query optimizations.
Implemented low-latency cache layer between Node.js backend and React client.
Containerized Python CV inference pipelines for reproducible one-command deployment.
Inspected server logs, diagnosed failing HTTP requests, and verified database state.
Managed sprint backlogs, REST API specifications, and cross-functional deliverables.
Tested release builds against user acceptance criteria to ensure zero regressions.
Autonomous 4-agent RAG engine with Presidio PII pseudonymization, AES-256-GCM cryptographic shredding, and EU AI Act record-keeping.
Standard enterprise LLM deployments expose corporations to severe privacy fines (up to €35 Million or 7% global turnover under GDPR) and hallucination liabilities. This system solves both by deploying a sovereign EU-bounded 4-agent team (Query Planner, Hybrid Retriever, Verifier, Synthesizer) that cross-examines every claim, sanitizes PII before vector indexing, and executes instant cryptographic shredding (GDPR Art. 17) upon user erasure requests without expensive vector re-indexing.
[Doc:Section].Under GDPR Article 33, any security breach involving personal data must be formally reported to the European supervisory authority within 72 hours of detection. All containment logs are immutably preserved in the audit ledger.
git clone https://github.com/Nehal-qadeer/eu-compliant-multiagent-rag.git && docker-compose up -dAn interactive educational tool using AI to bridge the communication gap through visual recognition.
This project was developed to help children with hearing impairments learn and identify objects in their surroundings through an engaging "Detection Game." Unlike standard detection tools, this is built as an interactive game where the AI scans the live webcam feed, identifies objects based on trained neural models, and displays the names in a large, easy-to-read font to provide immediate visual reinforcement for users who cannot rely on audio cues.
pip install opencv-python ultralytics openvino pyqt6Spatial room construction web application combining a 60 FPS HTML5 Canvas coordinate engine with Google Gemini Multimodal Vision AI.
Mastering German grammatical genders (der, die, das) and spatial room vocabulary is notoriously unintuitive through traditional rote memorization. This system bridges the gap by combining embodied 2D spatial placement on a 60 FPS HTML5 Canvas coordinate grid with real-time Google Gemini Multimodal Vision. Learners drag items into their virtual floor plan or scan real physical objects with their camera, receiving instant visual gender color reinforcement, contextual bilingual example sentences, and native Web Speech pronunciation.
git clone https://github.com/Nehal-qadeer/Virtual-Room-Builder-German-Game.git && npm run devPython and Selenium scrapers deployed as cloud Apify Actors with dynamic pagination and anti-bot bypass. Automated Make.com JSON validation transforms messy payloads before direct ingestion into PostgreSQL relational databases.
Dual systems: A 4-step Zapier automation with OAuth webhooks (Forms ➔ Sheets ➔ Calendar ➔ Gmail) combined with a full-stack real-time Satellite Tracker built in Node.js, React, and Redis as the shared distributed state and caching layer.
Focus: Distributed Systems, Software Engineering, Cloud Architectures.
Core coursework: Algorithms, Data Structures, Object-Oriented Programming (C++ & Python), Relational Database Systems.