Onpoint Insights· FEATURED
AI Engineer
Researched, architected, and shipped a government compliant automated medical coding pipeline as an AI Engineer for a healthcare client.
7.66%
Coding error rate, well below the 10–20% industry standard
3×
More accurate than a standard baseline
$8.4B
Market we are targeting in medical coding
01
Delivered an automated medical coding pipeline in 2 months, owning the project end to end as an AI Engineer.
Led end to end R&D as an AI Engineer for a medical coding pipeline, delivering a production ready, government compliant product in two months. Built the solution for a market projected to reach $8.4B by 2033, helping position the client at the forefront of medical coding automation.
02
Identified a flawed LLM baseline sitting below 70% accuracy before it became the foundation of the entire system.
Researched automated ICD 10 prediction approaches, identifying that a vanilla LLM baseline failed to exceed 70% coding accuracy and was not viable for clinical use. Established a retrieval based architecture as the team's technical foundation, preventing the team from building on a flawed baseline.
03
Built a retrieval system grounded in government sources, achieving 7.66% error rate vs a 10–20% industry baseline.
Integrated three official government CMS coding sources into the pipeline's retrieval layer, grounding every prediction in clinically and legally accepted standards. Achieved an error rate under 7.66%, significantly outperforming the 10–20% error rate typical of manual medical coding.
04
Designed a two stage AI pipeline that tripled prediction accuracy over a standard LLM approach.
Designed a two stage AI pipeline that progressively narrows ICD 10 candidates before classification, outperforming standard LLM baselines by over 3× in prediction accuracy. The staged architecture mirrors expert coding workflows, making the system both accurate and interpretable.
05
Embedded Medicaid and Medicare compliance rules directly into the pipeline to handle government payer requirements.
Incorporated Medicaid and Medicare compliance rules directly into the automated coding pipeline, ensuring outputs meet government payer requirements at the architecture level. This targets the $1B+ annual loss the healthcare industry incurs from coding errors and non compliant claims.
PythonRAGVector DatabasesLLMsICD 10CPTRetrieval ArchitectureMedical CodingHealthcare AIGovernment Compliance