Swaminathan Chellappa
RAG pipelines, LLM guardrails, and evaluation frameworks at USC's DILL Lab, including a HIPAA-compliant annotation platform now adopted by an external research organization. Previously shipped features to 23M+ users at Wells Fargo. MS Computer Science, USC '26.
SECTOR 1
Projects
SECTOR 2
Experience
Graduate Research Assistant · USC DILL Lab
Dec 2025 – May 2026
- ›"Live Codebook" system turning annotator corrections into LLM prompt rules: F1 up ~30% over zero-shot, annotation time from 3 hours to 12 minutes
- ›HIPAA-compliant local anonymization (87% entity detection); deployed across 4 GPUs on USC's HPC cluster, adopted by an external org
AI software engineering intern · koderAI
Jun – Aug 2025
- ›LLM benchmarking and routing framework across 32 models: production accuracy up ~26%, API retries down ~35%
- ›Guardrail endpoints (content + model-based filtering) securing internal AI workflows against prompt injection
Software engineer · Wells Fargo
Aug 2022 – Jun 2024
- ›Shipped BillPay and Alerts modules on the mobile platform serving 23M+ users; 100% WCAG accessibility compliance
- ›Sentiment dashboard over 10M+ app reviews, contributing to the app's 4.6 → 4.8 star rating
Machine learning intern · DataWeave
Nov 2021 – Jul 2022
- ›AWS S3 + MySQL inference pipeline with caching and batch orchestration: latency down 98% at ~100K records/day
WRITE-UPS
What On-Prem Deployment Taught Me About LLM Serving
How a privacy constraint walked me through three serving architectures, and a proof of concept that took the same four GPUs from 5 concurrent users to 30.
The Knobs Nobody Asked For: Learning to Build for the Customer, Not the Engineer
What pilot sessions with real researchers taught me about the difference between an interface an engineer wants and one a customer needs.
PIT WALL
Skills
AI/ML
Languages & frameworks
Infra & data
Testing & quality