Muhammad Murtaza Murtaza

About

Built a production LLM fine-tuning pipeline for domain-specific code generation at Pigeonbyte Technologies, including automated collection and rigorous quality filtering of 10M+ code samples (AST validation, sandbox execution/testing, deduplication, drift monitoring, and human-in-the-loop review). Also implemented end-to-end ML orchestration in Apache Airflow with data quality gates, dataset versioning in S3, benchmarking, and automated model promotion, and has a reliability-first approach to agent/workflow design.

Experience

Education

Key Strengths

• Built and deployed an end-to-end LLM fine-tuning pipeline for domain-specific code generation using 10M+ code samples
• Deep expertise in large-scale data quality management (AST syntax checks, sandbox execution, dedup hashing, complexity scoring, drift monitoring, human-in-loop review)
• Production-grade ML orchestration with Apache Airflow (DAGs spanning ingestion→training→evaluation→deployment, quality gates, alerts, dataset versioning in S3)
• Reliability-focused agent/workflow design (modular components, deterministic/reproducible runs, confidence-based fallbacks, circuit breakers)
• Able to translate AI capabilities/limits to non-technical stakeholders and manage expectations using measurable metrics (e.g., 7/10 to 8/10 working code generation)