Developed a remediation agent that reads ConnectWise tickets, links them to devices in a ConnectWise-compatible RMM, and plans safe script runs using GPT-5 via the OpenAI API on AWS Lambda with Python 3, integrating PostgreSQL, MongoDB, Redis, and AWS Secrets Manager for secure, low-touch operation.
Designed a 20-step microservice workflow for orchestration, device mapping, SOP compliance checks, and real-time script monitoring with intelligent retry logic and GPT-5 ticket closability analysis to auto-close resolved tickets and shrink time-to-resolution.
Built an AI agent for managed service providers that acts as an ongoing issue detector by grouping related tickets with HDBSCAN, NumPy, Pandas, and scikit-learn, helping teams spot emerging problems across customers.
Shipped an embedding-based classification service using cosine similarity and vector normalization to auto-assign new tickets to the right issue cluster in milliseconds, reducing manual triage and misrouting.
Created a natural language processing (NLP) ticket triage pipeline with a novel energy-based sanitization step that cleans and normalizes noisy tickets, improving downstream identification and automated fix suggestions.
Architected the production system on AWS with Lambda, SQS FIFO, EventBridge, API Gateway, Cognito, FastAPI, and MongoDB Atlas to support real-time ticket processing and cluster management.
Built a modular data engine with MongoDB aggregation pipelines, batch processing, and connection pooling so clustering and classification stay fast under high ticket volume.
Implemented a content filtering layer with fuzzy logic, regex, and rule-based checks to remove spam and low-quality tickets before they reach the agents.
Set up structured JSON logging, CloudWatch monitoring, CloudFormation infrastructure as code, S3 artifact storage, and CI/CD pipelines to support reliable deployments and production debugging.