I build autonomous AI systems that actually work in production. As an Agentic AI Engineer with 5+ years of experience, I specialize in designing multi-agent architectures that move beyond chatbots into real, measurable business automation.
Boston, Massachusetts, United States
Designed and implemented multi-agent research prototype using LangGraph for automated literature review and synthesis —
agents autonomously retrieve papers, extract key findings, identify contradictions, and generate structured summaries with
citation provenance.
Built experimental RAG evaluation framework benchmarking chunking strategies (fixed-size, semantic, document-aware),
embedding models (OpenAI, Cohere, BGE), and reranking approaches across 5 domain-specific corpora; published internal findings
on optimal retrieval configurations achieving 89% recall@10.
Developed Bedrock-integrated agentic pipeline for research data annotation, using structured Pydantic outputs and human-in-the-
loop validation to generate high-quality labeled datasets — reducing annotation time by 50% compared to manual labeling
workflows.
Implemented MCP (Model Context Protocol) server for connecting LLM agents to university research databases, enabling tool-use
grounded retrieval across 200K+ academic documents with semantic search and metadata filtering.
Built evaluation harness for measuring agent reliability: task completion rates, hallucination detection, tool-call accuracy, and
reasoning chain quality across multi-step workflows — used to benchmark 4 LLM providers systematically.
Boston, Massachusetts, United States
Architected production-grade agentic AI platform using LangGraph multi-agent workflows (specialist/reviewer/executor pattern),
enabling autonomous multi-step task planning and self-critique loops that reduced manual intervention by 65% across enterprise AI
pipelines.
Implemented robust RAG pipelines with hybrid search (dense + sparse), custom chunking/indexing strategies, and Pinecone vector
database integrated with Amazon Bedrock (Claude, Titan), achieving 92% retrieval accuracy and citation-first structured outputs for
semantic search applications.
Designed and enforced LLM guardrails including prompt/policy templates, structured JSON/Pydantic outputs, and allow/deny tool
lists to ensure auditability, safety, and reliability across 10,000+ daily AI inference requests.
Built serverless agentic backend on AWS Lambda, Step Functions, and EventBridge; instrumented agents with CloudWatch metrics,
spans, and traces for full observability and audit trails — maintaining 99.9% uptime.
Integrated multiple LLM providers (OpenAI, Anthropic Claude, Titan) through unified API architecture enabling seamless model
switching, reducing LLM inference costs by 30% while improving response quality.
Engineered CI/CD pipelines with GitHub Actions and Terraform for automated deployment, reducing deployment cycle from 4 hours
to 15 minutes; developed ReactJS monitoring dashboard improving operational visibility by 40%.
Pune, Maharashtra, India
Engineered cloud-native microservices on AWS (EKS, Lambda, RDS) processing 50,000+ daily transactions for enterprise
healthcare billing platform; implemented event-driven architecture with EventBridge and Step Functions reducing system latency
by 35%.
Developed NLP-based data extraction pipeline using spaCy and rule-based classifiers to process unstructured medical billing
documents (EOBs, remittance advices), achieving 88% field-level extraction accuracy and reducing manual data entry by 45%.
Built anomaly detection system for billing fraud identification using isolation forests and engineered features (claim frequency,
provider deviation scores, billing code patterns), flagging 3,200+ suspicious claims monthly with 91% precision.
Designed RESTful APIs using Java/Spring Boot integrated with PostgreSQL RDS supporting 99.99% availability SLA; built ReactJS
frontends achieving 25% improvement in page load performance.
Established automated testing frameworks achieving 85% code coverage, reducing production defects by 40%; delivered sprint
features 20% ahead of schedule collaborating with 12+ engineers and data scientists.
Pune, Maharashtra, India
Developed 20+ RESTful APIs for trading platform integration, processing $5M+ in daily transaction volume with sub-200ms response times
Created responsive web applications using Angular and JAVA, Spring Boo, serving 75K+ concurrent users during peak trading hours
Implemented automated testing frameworks using JUnit, Mockito, and Cucumber, achieving 92% code coverage and reducing production defects by 40%
Collaborated in Agile environment delivering 25+ features across 8 sprint cycles, consistently meeting 95% of committed story points
Designed UML system diagrams for 5 major system integrations, improving development team understanding and reducing implementation time by 30%
Pune
Architected and implemented identity management services supporting 15M+ monthly active users, reducing authentication failures by 42% while improving response times by 65%
Led migration of legacy authentication systems to scalable microservices architecture using Python and Kubernetes, resulting in 99.99% availability. Designed and deployed CI/CD pipelines with GitLab and Jenkins, reducing deployment time from 2 hours to 12 minutes and enabling 60+ weekly deployments
Implemented Infrastructure as Code using Terraform to provision and manage cloud resources across AWS and GCP, reducing infrastructure management time by 75%. Created comprehensive automated test suites that increased code coverage from 72% to 94%, decreasing production incidents by 61%
Mentored 4 junior engineers on DevOps best practices and security-focused development, resulting in a 35% improvement in code quality metrics
Education
Northeastern University
Master's degree
Indian Institute of Science (IISc)
Postgraduate Degree
Manipal Institute of Technology