Architect/Software Engineer

• Architected and built a high-performance, horizontally scalable NoSQL database in Go, leveraging Direct I/O B-Tree/Hash Map structures and Redis caching to achieve ACID compliance without third-party database dependencies

• Engineered the core Transactional Vector Database (SOP AI) module, combining B-Tree storage with Cosine Similarity search to enable ACID-compliant vector operations and real-time deduplication, overcoming the eventual-consistency issues of standard vector stores

• Invented a novel "Ideal Random Sampling" indexing algorithm using dense integer maps to generate mathematically representative search centroids, ensuring high-quality, unbiased clustering performance on large, skewed datasets

• Implemented a "Self-Healing" Index with background rebalancing and real-time distribution tracking, ensuring the vector store maintains optimal O(logN) search performance at terabyte scale

• Designed the cost-effective "Doctor & Nurse" AI Orchestration Pattern—a dual-agent architecture that leverages lightweight local LLMs (Ollama) for intent classification and the SOP Vector Store for deep, transactional knowledge retrieval

• Developed a Zero-Code Agent Framework featuring JSON configuration and automated ETL pipelines to cut expert system deployment time from weeks to hours, achieving 100% data freshness

• Built a hierarchical AI Policy Engine for Responsible AI, enforcing both global and local guardrails (e.g., PII filtering, safety checks) directly within the retrieval pipeline, enhancing system governance

• Deployed a robust, fault-tolerant replication layer using Reed-Solomon Erasure Coding and highly concurrent Go routines/channels, ensuring high durability and low-latency synchronization across the cluster

• Increased adoption by developing Python wrappers (via CGO/C hooks) for the core store API, enabling seamless integration for Python developers and leading to a publication on PyPI

Senior Software Engineer (Golang/Java/Postgres/Kubernetes)

• Authored the Jobs REST API (Java/PostgresDB) using DDD/SOLID principles, employing lightweight threads in JDK23 for a highly threaded Kubernetes jobs processing pipeline.

• Provisioned and configured a bare-metal Kubernetes cluster including a high-performance, distributed MinIO (S3 compatible) cluster with dual-node redundancy (Helm chart).

• Led R&D on future-facing components, including OAuth2-based authentication and a RESTful, cached CDN based on the SOP project written in Golang/Swagger.

Consultant (Golang/Java/GoogleSQL/CEL)

• Delivered scalable MicroServices/CEL Rules and backend frameworks in Golang/CEL/GoogleSQL for Enterprise Infrastructure Protection, enforcing security checks across real-time and high-volume background data processing.

• Contributed to Google’s standardized logging design pattern by implementing Sawmill logging across several JVM/Java RPCs.

• Optimized implementations using Golang channels and Go-routines to take advantage of high concurrency for scalable query workloads.

Sr Consultant (Golang/MongoDB/Kubernetes)

• Designed and solely implemented the Member Search Hydrator application, achieving 40 million Member records’ hydration in 15 hours with low latency and non-disruptive data sync.

• Enhanced the Kafka library with a multi-threaded work-pool based intake for higher speed message processing, leading to improved team throughput.

• Authored the IMT-Orchestrator micro-service, integrating three APIs and applying business logic, utilizing a multi-threaded worker library (Go channels/routines) for high-speed request processing.

Sr. Software Engineer (Java/Golang/AWS/Redis/ElasticSearch/Kafka)

• Authored the GOLANG Pages REST web service and data management module, providing Geo-spatial data-aware RadioStations metadata lookup from Redis/ElasticSearch with built-in cache failure handling.

• Led the migration of the proprietary search engine to a SOLR 8 based solution, identifying and fixing performance bottlenecks that caused unacceptable slowdowns.

• Increased overall linking throughput from 250K to 1M/day by porting the Music Linker cluster from Lucene to an ElasticSearch AWS cluster and designing an efficient document schema that reduced indexes from 4 to 1.

• Optimized the ScanAndMatch application to sustain performance and increase ingestion rate from 250K up to 800K processed Album submits daily by resolving bottlenecks in the data pipeline.

• Designed and implemented an ultra-fast, highly parallelized AWS cloud migration utility capable of performing terabytes file copy operations per hour.