AI Engineer

• Leading development of an AI agent delivery platform that understands and processes prospect emails, enabling direct client use for automated lead qualification and response generation.
• Architected end-to-end agent pipeline using LangGraph with specialized agents for email parsing, intent classification, sentiment analysis, and response drafting.
• Implemented prompt engineering strategies and guardrail mechanisms to ensure reliable, safe, and contextually appropriate agent responses.
• Deployed agentic workflows on Azure infrastructure with Kubernetes orchestration, ensuring high availability and sub-second response times.
• Integrated vector search capabilities for retrieval-augmented generation, enabling the agent to access historical email patterns and company knowledge bases.
• Collaborated directly with client stakeholders to refine agent behavior, incorporating feedback loops and continuous improvement mechanisms.

• Embedded with a Fortune 500 enterprise client to architect and deploy a 9-node LangGraph intelligence pipeline implementing the Leiden Intelligence Cycle, automating query decomposition and structured reporting, reducing analyst research time by 55%.
• Orchestrated complex Azure Data Factory and Azure Machine Learning pipelines, debugging and resolving critical YAML syntax and parameter mapping issues to ensure zero-downtime CI/CD workflows and reliable model deployment.
• Owned end-to-end delivery of a multi-tenant Pinecone vector database backend with delta-sync ingestion, Parent-Child chunking, and Hybrid Search (Dense + Sparse), improving retrieval F1-score by 30% for 10,000+ documents.
• Led bi-weekly technical workshops with client VPs and Directors, translating LLM and infrastructure decisions into business ROI and securing contract expansion.
• Hardened AI deployment with enterprise security layers (prompt-injection guardrails, PII redaction); passed internal security audits with zero critical findings and maintained 99.9% uptime over 12 months.
• Optimized inference costs via PEFT (LoRA/QLoRA), 4-bit quantization, and vLLM serving, reducing GPU inference spend by 45% while maintaining output quality within client SLA thresholds.

• Engineered the transition from a monolithic internal dashboard to a decoupled microservices architecture on GCP; implemented a centralized API Gateway handling 10,000+ concurrent requests with sub-100ms p95 latency.• Enforced rigorous Git version control and CI/CD protocols, directing code pushes to specific, isolated feature branches to ensure clean pipeline triggers and prevent main-branch regressions.
• Designed a multi-tier caching layer using Redis (Pub/Sub + Distributed Caching), reducing database read contention by 60% and optimizing dashboard Time-To-First-Byte (TTFB) for 2,000+ global users.
• Architected blue-green deployment pipelines using Docker and GitHub Actions, achieving zero-downtime releases for 20+ monthly deployments with automated rollback on failure detection.

• Developed a full-stack analytics dashboard for a client processing 1M+ data points, implementing both backend APIs and responsive frontend components with direct client collaboration.
• Architected data visualization pipeline with real-time updates, enabling clients to monitor key metrics and make data-driven decisions.
• Optimized database queries and implemented pagination strategies, reducing data retrieval time by 35% for large datasets.
• Designed and implemented RESTful APIs using Node.js/Express, connecting frontend React components with PostgreSQL backend.
• Engaged directly with client stakeholders through weekly standups to gather requirements, demonstrate progress, and incorporate feedback iteratively.
• Implemented role-based access control and secure authentication for different user types accessing the dashboard.

Tech: CrewAI, FastAPI, PostgreSQL, Redis, React

• Architected a production-ready multi-agent swarm using CrewAI/AutoGen with 6 specialized agents (Supervisor, Researcher, Budgeter, Logistics, Activity, Executor) for autonomous travel planning.
• Designed parallel processing architecture where Supervisor decomposes tasks and triggers Research, Budget, and Logistics agents concurrently, reducing end-to-end response time by 60%.
• Built tool integration for each agent using Google Places API, OpenWeatherMap, Amadeus/Skyscanner API, and Booking.com API for real-time data retrieval.
• Implemented state management and context sharing between agents, enabling seamless handoff of constraints, budgets, and discovered information across the swarm.
• Integrated RAG system for citation generation, ensuring every recommendation includes document references (doc id, chunk id) for transparency and trust.
• Enforced tenant isolation with pre-retrieval org id filters, preventing cross-organization data leakage in the RAG-powered recommendation engine.
• Deployed rate limiting and usage metering to enforce subscription quotas across 100+ organizations with tiered plans.
• Designed real-time progress tracking with WebSocket updates, allowing users to see agent swarm progress and intermediate decisions during planning.

• Architected an enterprise-grade multi-tenant RAG platform with SuperAdmin and Organization-level RBAC (Org Admin, End User) supporting 20+ document formats (PDF, DOCX, PPTX, XLSX, images, audio via Whisper, Confluence, SharePoint).
• Designed tenant isolation architecture with strict row-level security — every query filtered by org id derived from JWT/OAuth2 token, preventing cross-tenant data leakage.
• Implemented pluggable parser/connector architecture with async ingestion pipeline (queue + worker pool) supporting 20+ formats and source connectors (Confluence, SharePoint, Google Drive, S3).
• Integrated hybrid search (BM25 + Dense Vector) with re-ranker for improved retrieval quality across 10,000+ documents.
• Built subscription management with tiered plans (Free/Tier1/Tier2/Enterprise) including usage metering, quota enforcement, and Stripe webhook integration for billing.
• Developed observability stack with structured logs, tracing per request, and cost dashboards per organization; implemented audit logs for GDPR-style compliance.
• Designed security layer including data encryption at rest/in transit, signed URLs for file access, MFA for admins, and SSO/SAML support for enterprise organizations.