Daven Chang

Software Engineering Intern

• 01Design a 12-component scoring engine ranking 7,000 CMBS deals via cosine similarity, log-distance, and hybrid Jaccard metrics.
• 02Integrated pgvector semantic search using 768-dim all-mpnet-base-v2 embeddings for natural-language bond discovery.
• 03Replaced a manual spreadsheet process with a FastAPI ETL ingestion microservice parsing six entity types into PostgreSQL.
• 04Built a data aggregation layer flattening 102 nested loan fields into a 35-field scoring vector with feature normalization.
• 05Implemented Redis caching for scoring queries against the 7,000 deal corpus, reducing response times to sub-second.

Software Engineering Intern

• 01Engineered an onboarding framework to parse documentation and map endpoints, cutting setup time by 93%.
• 02Slashed regression detection from 4hr to under 5min with a CI pipeline linking git merges to Grafana performance reports.
• 03Orchestrated Kubernetes to automate distributed stress-tests against Java Spring Boot microservices at 20,000+ TPS.
• 04Integrated Terraform into Jenkins CI/CD to provision ephemeral AWS testing suites, including Cassandra clusters.

Technical Lead (Client: Booz Allen Hamilton)

• 01Led 12 developers in shipping a React/Express and LlamaIndex RAG platform, slashing policy review times to 15 mins.
• 02Engineered a high-throughput, serverless FastAPI ingestion framework via API Gateway, AWS Lambda, and AWS SQS.
• 03Designed a PostgreSQL schema modeling 150+ intricate relationships for the enterprise policy management system.
• 04Architected a multi-AZ RDS cluster across three availability zones with a custom EC2 NAT Gateway for database failover.

Machine Learning Researcher

• 01Authored a departmental publication achieving a 22% R2 improvement via a hybrid PyTorch LSTM-DNN architecture.
• 02Built Pandas/NumPy feature pipelines modeling lobbying and ESG risk dynamics that powered the architecture above.
• 03Designed a parallelized Python/Selenium ETL pipeline ingesting large-scale corporate filing datasets for training.