Research Focus Areas
Multimodal Architectures
Optimization
Interpretability
High Performance Computing
Scroll
Experience
October 2024 - Present
October 2024 - Present
Payment Prediction System: Deployed ML model improving successful payment rates from 47% to 59%, directly impacting revenue through data-driven lead prioritization
Real-Time Voice AI Platform: Architected production voice agent system reducing response latency by 85% (from 5-7s to sub-second) through model optimization and streaming techniques
Automated Quality Assurance: Built serverless architecture processing 99% of call transcripts with automated compliance scoring and real-time alerting
Multi-Objective Optimization: Developed optimization system achieving 3-5% improvement in plan recommendations while reducing training time by 21x through evolutionary algorithms and differentiable ranking formulations
RAG Infrastructure: Created production retrieval system for insurance documentation with sub-second response times (p95 < 800ms) and 0.91 MRR@10
AI Agent Integration: Built MCP server connecting Neo4j knowledge graphs with internal APIs, enabling AI agents to leverage company data for enhanced decision making
ML Infrastructure: Designed feature store on Databricks with automated quality monitoring and real-time performance tracking
Technical Leadership: Chair AI tooling committee and facilitate biweekly ML research presentations
May 2021 - Present
May 2021 - Present
Transformer Architecture for Clinical AI: Developed custom multi-modal transformer from scratch, improving glaucoma prediction from 0.74 to 0.97 AUC through innovations in handling irregular temporal data
Uncertainty Quantification: Implemented probabilistic modeling for clinical decision support with counterfactual analysis and trajectory forecasting
Clinical NLP System: Built classification system for medical notes using fine-tuned LLMs, achieving 0.96 AUC and deployed on Azure for PHI-compliant production use
Computer Vision Pipeline: Fine-tuned foundation models for automated medical image segmentation (Dice 0.88) and developed novel algorithms for clinical measurements
Research Data Infrastructure: Led infrastructure managing 10TB+ medical data, enabling multiple concurrent research projects and publications
May 2019 - December 2021
May 2019 - December 2021
Quantitative Finance ML: Led team developing neural networks for S&P 500 return forecasting, identifying key predictive features from 94 stock characteristics
Neural Network Optimization: Developed mathematical formulations for network analysis, discovering that pruned networks can achieve higher adversarial robustness while maintaining accuracy
Selected Publications
INDEX: 12 ITEMSDetecting glaucoma worsening using optical coherence tomography derived visual field estimates
Scientific Reports (2025) • Pham, A. T., Bradley, C., Hou, K., Herbert, P., & Yohannan, J.
The Impact of Achieving Target Intraocular Pressure on Glaucomatous Retinal Nerve Fiber Layer Thinning in a Treated Clinical Population
American Journal of Ophthalmology (2024) • Pham, A. T., Bradley, C., Hou, K., Herbert, P., Boland, M. V., Ramulu, P. Y., & Yohannan, J.
Deep learning-based identification of eyes at risk for glaucoma surgery
Scientific Reports (2024) • Wang, R., Bradley, C., Herbert, P., Hou, K., Ramulu, P., Breininger, K., Unberath, M., & Yohannan, J.
Take a Closer Look
My background combines academic rigor with production engineering. You can read the full CV, or ask the AI assistant about specific implementation details.