Engineering the cognitive layer of an agentic AI system targeting logistics industry complexity — leveraging LangGraph for stateful orchestration and LangChain toolchains to enable autonomous decision-making, decoupled from downstream execution systems.

Scaled the patent litigation analytics and intelligence platform to 200K+ monthly visitors by leading full-cycle product development — owning architectural decisions, data pipeline reliability, and cross-functional feature delivery across the IP industry.

Built an LLM-powered natural language search system with schema-aware SQL generation and constraint validation — enabling non-technical users to query complex patent datasets with zero hallucination risk in production.

Led end-to-end development of a global drug and patent monitoring platform, designing multi-tenant architecture with RBAC to enable pharmaceutical clients to track all FDA-approved drugs and associated patent landscapes across global jurisdictions.

Engineered asynchronous ingestion pipelines and a queue-based alert system to monitor 1M+ patents in real time, delivering zero-message-loss portfolio notifications while reducing data processing latency by 40% under high-throughput conditions.

Integrated Claude and ChatGPT into patent classification pipelines via prompt engineering, automating categorization by type and drug portfolio mapping — achieving 90%+ accuracy across multilingual datasets.

Identified a critical gap in patent litigation analytics and drove ground-up development of the analytics platform — architecting the system across a FastAPI / PostgreSQL backend, Nuxt 3 frontend, and AWS infrastructure, scaling it to 100K+ monthly users.

Built a serverless data ingestion pipeline on AWS Lambda to automate scheduled ETL of patent datasets, achieving over 20x improvement in processing throughput and enabling daily refresh cycles with no manual intervention.

Designed and shipped Python-based automation tools to replace analyst-operated manual workflows, cutting processing time and reducing error rates by more than 90% — directly improving downstream data reliability for product features.

Designed and optimized PostgreSQL schemas and queries for large-scale patent datasets, applying advanced indexing and query tuning techniques that delivered up to 40% improvement in query performance across high-volume analytics workloads.

Built a CRUD-based Feedback REST API using Node.js, handling end-to-end user feedback submission, processing, and persistent storage in the database.

Queried and analyzed 1M+ retail records using MongoDB aggregation pipelines to identify purchasing trends and customer segments.