Sai Krishna Sriram

About

AI/ML engineer who shipped a production blood-test report understanding and personalized supplement recommendation product, using a LangGraph multi-agent pipeline on AWS serverless with OCR via Bedrock and RAG over vetted clinical research. Also built end-to-end recommender system pipelines at ASANTe using Airflow (ingestion, embeddings/features, training, registry, batch scoring/monitoring) with KPI reporting to Tableau, with a strong focus on safety, evaluation, and measurable reliability.

Experience

Education

Key Strengths

• Built and deployed a production multi-agent blood report analysis + supplement recommendation system on serverless AWS
• Designed robust orchestration for AI/ML workflows using AWS Step Functions (Distributed Map) and Apache Airflow DAGs
• Improved factual accuracy and reduced hallucinations via RAG with clinical research vectors (35% accuracy increase, 20% hallucination reduction)
• Strong approach to reliability: golden test sets, schema validation, groundedness checks, strict JSON/Pydantic, timeouts, max-step limits, fallbacks, and run auditing
• Expertise in messy real-world data standardization across heterogeneous lab report formats and inconsistent test naming
• Ability to translate clinician/product safety requirements into guardrails, rules, UX constraints, and measurable evaluation metrics

Reference Highlights

• Takes clear ownership and drives work to production readiness
• Reliable and delivers tasks on time
• Strong technical problem-solving
• Proactive in cross-team collaboration
• Handles messy, changing data effectively
• Improves system performance under high traffic
• Structured, clear communicator
• Explains complex systems in simple terms (starts with the 'why')
• Patient in answering questions and aligning stakeholders
• Effectively incorporates stakeholder feedback into product/system improvements