Senior Software Engineer

Architected production-ready, multi-tenant knowledge graph APIs using Python, FastAPI, and Neo4j. Built graph traversal, BM25 full-text search, relationship suggestions, KPI hierarchy modeling, RBAC, tenant isolation, background import processing, and OpenAPI documentation for enterprise asset-management and analytics products.

Delivered LLM and RAG systems using Azure OpenAI, OpenAI GPT, LangChain, LangGraph, Hugging Face Transformers, Milvus, Ollama, and internal vector databases. Developed AI copilots, chatbots, document ingestion and embedding pipelines, forecasting models, video analytics pipelines, and fraud-detection solutions.

Improved platform reliability and cost efficiency through Kubernetes autoscaling, client-side load balancing, Prometheus and Grafana observability, CI/CD, and comprehensive automated testing with RSpec, Pytest, Cypress, Behave, and Tavern. Mentored engineers and led cross-functional architecture reviews with product, design, and SRE teams.

Designed a graph-based maritime trajectory analytics platform that ingested AIS data, processed vessel movements, and stored semantic and geospatial relationships in Neo4j AuraDB. Created algorithms for trajectory chunking, next-voyage links, navigation attribute categorization, relationship modeling, and risk-property tagging.

Built a petabyte-scale, serverless streaming ingestion pipeline sustaining 12 GB/s throughput and processing 1 PB of data and 1.04 trillion records in 24 hours. Developed geospatial utilities, embedding-based similarity search, anomaly detection, and a full-stack mapping application using OpenStreetMap, cosine-similarity queries, and spatial indexing.

Developed Java and Go backend services for real-time robot task allocation, inventory movement, and warehouse fulfillment workflows. Built REST APIs, monitoring dashboards, and event-driven microservices for robot telemetry using React, Flask, Kafka, AWS Kinesis, and Grafana.

Created Python and WebGL simulation tools to visualize robot traffic, collision avoidance, and throughput bottlenecks. Optimized message queues and task-scheduling algorithms, partnered with hardware and mechanical engineering teams on A/B tests, and contributed to improved incident triage and warehouse pick efficiency.