Project 001 · Spark Growth · 2025
Creative Intelligence Agent
Architected a full-stack LLM-powered analytics system enabling natural language querying of social media creative performance data (Instagram, Facebook, LinkedIn) for DTC brand clients. Built a normalized Cloud SQL schema for creative metrics, a FastAPI + Claude API tool-use backend, and explored Vertex AI vector embeddings for semantic retrieval and text-to-SQL latency optimization.
Claude APIFastAPICloud SQLVertex AIGCPRAG
Project 002 · Spark Growth · 2025
Calendly → CRM Webhook Pipeline
Built a webhook-driven automation pipeline on GCP Cloud Run integrating Calendly with Productive CRM — automating company/contact creation, deal generation with industry mapping, and duplicate detection. Handles 50+ bookings/week end-to-end with zero manual intervention.
PythonCloud RunGCPWebhooksREST APICRM
Project 003 · Spark Growth · 2025
Multi-user Data Export App
Developed a multi-user data export application on GCP Cloud Run with FastAPI, featuring Google OAuth 2.0 authentication, role-based access control, and seamless exports to Google Sheets and CSV — enabling the team to self-serve reporting without engineering support.
FastAPICloud RunGCPOAuth 2.0RBACGoogle Sheets
Project 004 · Apr – May 2025
Fraud Detection MLOps Pipeline
Designed and containerized a FastAPI microservice with MLflow-registered models for real-time and batch fraud inference. Orchestrated a multi-service architecture via Docker Compose, enabling rapid local simulation of production ML pipelines. Implemented model versioning and deployment registry following reproducible MLOps best practices.
PythonFastAPIMLflowDockerDocker Compose
Project 005 · Jan – Apr 2025
MBTI Multimodal Deep Learning
Developed a hybrid BERT + CNN model for multimodal personality inference. Applied data augmentation and prompt tuning to address class imbalance. Reduced training time by 60% leveraging CUDA-enabled GPUs (RTX 5090) and optimized batch loading pipelines.
PyTorchBERTCNNCUDAPythonNLP
Project 006 · Jan – Apr 2025
Course Enrollment Platform
Built a scalable SaaS-style platform supporting real-time enrollment for 500+ courses with a Java RESTful API and interactive frontend. Implemented RBAC for 3 user roles, enabled real-time availability via WebSockets (reducing booking conflicts by 40%), and optimized SQL queries by 45% through indexing and join restructuring.
JavaJavaScriptSQLWebSocketsREST APIAgile
Project 007 · Sep – Dec 2024
Resource Allocation Optimizer
Led a team to design a demand-driven scheduling system across multiple customer zones. Extracted spatial-temporal demand patterns and built forecasting models (Random Forest) with constrained optimization (cvxpy) to create order allocation plans under limited capacity — improving service level by 34%.
PythonScikit-learncvxpyRandom ForestExcel
Project 008 · Jan – Apr 2025
System Simulation & Capacity Planning
Developed discrete-event simulations in SimPy to model system capacity, routing, and cycle times under fluctuating demand. Performed bottleneck analysis using Pandas/NumPy, and optimized resource allocation strategies with SciPy — reducing overall system operating cost by 60% through scenario-based planning.
PythonSimPySciPyPandasNumPyExcel Solver
