Partner closely with attorneys and operations to gather requirements and redesign the client intake process, transforming manual workflows into a fully automated, API-driven pipeline.

Architected an intake automation pipeline integrating Google Forms, Custom OCR processing, and Clio API to automate data population and validation.

Implemented automation tooling including visa expiry reminders, LLM-generated consultation summaries, AWS S3 disaster recovery backups, and uptime monitoring for production reliability.

Currently expanding AI-driven document classification and structured data extraction to further reduce manual case processing.

Developed a contact deduplication engine using fuzzy matching and normalized email/phone validation, resolving 4,500+ corrupted or duplicate records.

Built and standardized a 5-channel (RGB, DSM, Flux) rooftop detection dataset using Google Solar API imagery, integrating YOLO-SAM with CVAT to semi-automate COCO annotations across 1,100+ images and improve labeling consistency.

Resolved critical data augmentation and patching issues using sliding window logic and dynamic padding, stabilizing the training pipeline and eliminating duplication errors.

Increased model accuracy from 82% to 95% by optimizing multi-channel normalization, augmentation strategies, and loss weighting, while supporting backend integration with a PHP Laravel platform to automate solar layout estimation and cost projections.

Built RESTful APIs using Java Spring Boot with role-based access control (RBAC) and multi-factor authentication (MFA).

Optimized backend performance to achieve sub-200ms response times using caching, async processing, and WebSocket-based real-time updates.

Designed and integrated an AI chatbot for FAQs and user support, reducing support queries by ~25%.

Contributed to full SDLC including requirements gathering, development, testing, and deployment.