Susmitha Dodla

• Designed and deployed a document intelligence platform enabling 30+ analysts to query 1,000+ research reports via metadata-aware chunking, semantic search, and RAG — cutting average document lookup time by ~40%. • Architected RAG pipelines integrating LLMs (GPT-4) with FAISS vector search and LangChain orchestration, reducing irrelevant retrievals by ~25% through hybrid retrieval tuning and prompt-template refinement. • Built hallucination evaluation and guardrail workflows across 500+ daily queries, using systematic failure analysis and prompt iteration to improve response consistency by ~35%. • Developed FastAPI inference services handling 500+ daily requests at sub-200ms latency via model selection, request batching, and response caching strategies. • Led AWS deployment (Lambda, EC2, S3) and CI/CD pipelines, improving end-to-end inference latency by ~30% through infrastructure tuning and efficient request handling. • Built monitoring dashboards tracking usage, response quality, and system health across 4 teams — reducing issue resolution time by ~20% and enabling data-driven product prioritization. • Partnered with product and data engineering teams to embed AI-driven solutions into daily analyst workflows, measurably reducing manual document review time and driving broad adoption.

• Built ML classification and risk identification models on structured and clinical datasets, improving predictive performance by 15–20% through iterative feature engineering, cross-validation, and error analysis. • Engineered end-to-end data preparation pipelines (Pandas, NumPy) to clean, standardize, and validate 5+ healthcare data sources covering 200K+ patient records, reducing data error rates by ~30%. • Applied NLP techniques — embeddings and text preprocessing — to extract high-signal features from unstructured clinical notes, directly improving downstream model quality. • Drove iterative model improvement cycles by diagnosing misclassification patterns and refining feature sets, delivering measurable robustness gains across multiple validation rounds. • Designed and deployed Flask inference APIs handling 1,000+ daily requests, integrating ML outputs into 3 internal systems across 2 product lines for analytics and clinical teams. • Built 10+ Tableau and Power BI dashboards communicating model outputs and operational metrics to 20+ stakeholders, compressing reporting cycles from weekly to near real-time. • Implemented data privacy and validation workflows aligned with healthcare data standards, ensuring secure, compliant handling of sensitive patient records. • Led code reviews for a 3-person ML sub-team, contributing to a ~20% reduction in post-release defects and improved delivery velocity across 6 sprint cycles.

Completed a Master’s degree in Computer Science from the University of Central Missouri with a strong academic foundation (GPA: 3.60). Gained hands-on academic and project-based exposure to data engineering, analytics, cloud computing, and DevOps fundamentals, including SQL, Python, AWS, CI/CD concepts, and system automation.