Nabeel Hussain

Impact in production

Now

Conovo Technologies, Lahore (Aug 2023 - Present)

Core

Python, Django, FastAPI, Flask, PostgreSQL, Redis, Celery

AI

RAG, LLM integration, multi-agent systems, Claude API, GPT-4, LangChain

Ops

AWS, Docker, Terraform, Jenkins CI/CD, Nginx

Degree

BS Software Engineering, GC University Faisalabad

Certs

Azure AI Apps and Agents Developer Associate, Azure Fundamentals, AWS Cloud Practitioner

• 24s→1sAPI response time after query, backend and similarity-search optimisation ArhamSoft
• 250+websites scraped with Airflow, Scrapy and Selenium Conovo
• -40%inference time with GPU acceleration 21Swap
• +30%performance from data processing and query tuning Conovo scraping, Hades
• +25%detection accuracy on handwritten envelopes EnvelopeAI

+30% performance

Hades

Tracks and filters sellers of fake products.

• Spearheaded the application and designed the backend from scratch, leading a team of three developers.
• Scheduled the scraper DAGs with Apache Airflow to run the scraping process efficiently.
• Solved scaling issues by optimising database queries, for a 30% performance boost.
• Automated scraper maintenance with a custom-trained LLM.

• Qwen2.5
• Claude 3.5
• Apache Airflow
• Scrapy
• PostgreSQL
• AWS EC2, S3

-40% inference time

21Swap and FaceFusion

AI image and video generation with deep learning face swapping.

• Led development of 21Swap, an AI-powered face-swapping system for images and video.
• Designed and optimised neural networks for high-fidelity, realistic transformations.
• Added GPU acceleration for real-time processing, cutting inference time by 40%.
• Built automated preprocessing and augmentation pipelines for model robustness and accuracy.
• At Conovo, worked on FaceFusion, optimising model performance for accuracy and efficiency.

• Python
• TensorFlow
• OpenCV
• Flask
• AWS GPU instances

OCR to classified text

Text extraction pipeline

Reads documents and sorts what they say.

• Designed and implemented a text extraction architecture with PaddleOCR for OCR.
• Classified the extracted text with SpaCy transformers, strengthening the data processing pipelines.

• PaddleOCR
• SpaCy
• Transformers
• Python

+25% accuracy

EnvelopeAI

Finds sender and receiver details on envelope images, handwriting included.

• Built a custom YOLOv8 detection model for the sender and receiver blocks.
• Iterative training on annotated data raised accuracy by 25% on handwritten text.
• Added OCR to extract the handwriting automatically.
• Trained on Google Colab for fast experiments, cutting model development time by 30%.

• YOLOv8
• PyTorch
• OpenCV
• OCR
• Google Colab

Stripe payments built in

AssetCurve

Secure, cost-effective, scalable IT asset management.

• Monitored IT hardware and users, with insight into equipment costs and streamlined recertifications.
• Managed company data, applications and their users securely.
• Built the backend on Django and DRF over AWS S3, RDS, CloudFront and EC2, with Stripe for payments.

• Django
• DRF
• PostgreSQL
• AWS S3, RDS, CloudFront, EC2
• Nginx
• Stripe