Muhammad F.

Key Skills

Software

Other

Top Subject Matter

Satellite imagery analysis and geospatial scene understanding

LLM integration and AI agent training/inference workflow

Knowledge graph modeling for AI recommendation

Top Data Types

Image

Top Task Types

Segmentation

Function Calling

Land Cover Classification

Freelancer Overview

GeoExtract v2 (Final Year Project). Core strengths include Other. Education includes Bachelor of Science, Dawood University of Engineering & Technology (2026). AI-training focus includes data types such as Geospatial, Tiled Imagery, and Computer Code and labeling workflows including Segmentation, Function Calling, and Land Cover Classification.

Labeling Experience

Movie Recommendation System

OtherLand Cover Classification

Created a graph-based movie recommendation system using entity modeling and relationship-driven similarity matching. Modeled actors, directors, genres, and movies as nodes and used relationship traversal to infer relevant recommendations. Delivered the application via a deployed web interface with CI/CD automation. • Modeled knowledge graph entities in Neo4j using Cypher queries • Used ACTED_IN, DIRECTED, and IN_GENRE relationships for similarity matching • Implemented recommendation logic through graph traversals and filtering • Deployed a Streamlit app with Azure CI/CD for end-user access

2025 - 2026

Omni-Agent

OtherFunction Calling

Built an AI agent that routes requests across multiple LLM providers using an automatic router and context compaction strategy. Implemented provider failover behaviors to support consistent generation during rate-limit events. Packaged the agent as a reusable CLI/npm tool with BYOK-style architecture. • Implemented multi-provider LLM routing and context compaction on model switch • Added rate-limit failover logic across Groq/Gemini/Cerebras/SambaNova/OpenRouter • Delivered agent functionality via a TypeScript CLI interface • Published the solution as an npm package (npx omnillm) for reuse

2025 - 2026

GeoExtract v2 (Final Year Project)

OtherSegmentation

Developed a multi-model AI assistant for satellite imagery analysis, combining object detection and visual question answering for geospatial tasks. Built a YOLO-based pipeline to identify and segment structures within imagery and paired it with land-use classification. Created an interactive interface to generate reports and exported results for downstream review and analysis. • Implemented YOLOv11x-seg workflows for building detection on satellite images • Used EfficientNet-B2 for land-use classification over image inputs • Integrated Qwen2.5-VL for visual Q&A to support interpretation of imagery • Produced exportable outputs (PNG/JSON/CSV/GeoJSON) for structured evaluation

2025 - 2026

Education

Work History