This paper reports on a large-scale experiment introducing and evaluating intelligent tutoring in an urban high school setting. The Pittsburgh Urban Mathematics Project (PUMP) developed an algebra curriculum focused on real-world mathematical analysis and computational tools. An intelligent tutor, PAT, was created to support this curriculum and integrated into 9th-grade algebra in three Pittsburgh schools. In the 1993-94 school year, students in experimental classes outperformed comparison classes by 15% on standardized tests and 100% on tests targeting PUMP objectives. This study shows that intelligent tutoring systems can be effectively scaled and implemented in real-world settings like urban high schools. The PAT tutor was developed through a collaboration between PUMP and Carnegie Mellon University. It is practical in its pedagogical focus, engaging students in real-world problem-solving using algebraic tools. PAT is also practical in being a fully functioning system used by over 500 high school students. The system combines content guidance from mathematics pedagogy experts with scientific support from ACT theory and cognitive tutoring technology. The PUMP curriculum uses real-world situations to make mathematics more meaningful and accessible. It emphasizes functional models represented in tables, graphs, and symbols. The curriculum aligns with NCTM recommendations, focusing on real-world problems, computer utilities, and mathematical communication. PAT supports this curriculum by providing computational tools and an organized curriculum of problem situations. It was developed with input from curriculum designers and classroom teachers. The tutor uses a cognitive model based on ACT theory, which includes production rules for generating solution steps and mis-steps. This model supports two student modeling techniques: model tracing and knowledge tracing. Model tracing monitors student progress through problem-solving, while knowledge tracing tracks learning from problem to problem. PAT provides timely feedback and individualized help, reducing student frustration and providing a sense of accomplishment. It also uses knowledge tracing to identify individual areas of difficulty and present targeted problems. The tutor's feedback and help messages are critical for cognitive and motivational benefits. The study compared the PUMP+PAT curriculum with a traditional curriculum. Results showed that PUMP+PAT students scored significantly higher on standardized tests and reform-oriented assessments. They outperformed comparison classes by 15% on standardized tests and 100% on reform objectives. These results indicate that intelligent tutoring systems can be effective in real-world settings, improving student performance without compromising basic skills.This paper reports on a large-scale experiment introducing and evaluating intelligent tutoring in an urban high school setting. The Pittsburgh Urban Mathematics Project (PUMP) developed an algebra curriculum focused on real-world mathematical analysis and computational tools. An intelligent tutor, PAT, was created to support this curriculum and integrated into 9th-grade algebra in three Pittsburgh schools. In the 1993-94 school year, students in experimental classes outperformed comparison classes by 15% on standardized tests and 100% on tests targeting PUMP objectives. This study shows that intelligent tutoring systems can be effectively scaled and implemented in real-world settings like urban high schools. The PAT tutor was developed through a collaboration between PUMP and Carnegie Mellon University. It is practical in its pedagogical focus, engaging students in real-world problem-solving using algebraic tools. PAT is also practical in being a fully functioning system used by over 500 high school students. The system combines content guidance from mathematics pedagogy experts with scientific support from ACT theory and cognitive tutoring technology. The PUMP curriculum uses real-world situations to make mathematics more meaningful and accessible. It emphasizes functional models represented in tables, graphs, and symbols. The curriculum aligns with NCTM recommendations, focusing on real-world problems, computer utilities, and mathematical communication. PAT supports this curriculum by providing computational tools and an organized curriculum of problem situations. It was developed with input from curriculum designers and classroom teachers. The tutor uses a cognitive model based on ACT theory, which includes production rules for generating solution steps and mis-steps. This model supports two student modeling techniques: model tracing and knowledge tracing. Model tracing monitors student progress through problem-solving, while knowledge tracing tracks learning from problem to problem. PAT provides timely feedback and individualized help, reducing student frustration and providing a sense of accomplishment. It also uses knowledge tracing to identify individual areas of difficulty and present targeted problems. The tutor's feedback and help messages are critical for cognitive and motivational benefits. The study compared the PUMP+PAT curriculum with a traditional curriculum. Results showed that PUMP+PAT students scored significantly higher on standardized tests and reform-oriented assessments. They outperformed comparison classes by 15% on standardized tests and 100% on reform objectives. These results indicate that intelligent tutoring systems can be effective in real-world settings, improving student performance without compromising basic skills.