Ishan Aggarwal

Architected an AI Interview Coach and Data Catalog using GPT/Claude and LangChain, delivering sub-200ms real-time voice coaching with multi-turn conversational memory and fault-tolerant sessions. Deployed multi-tenant LLM agents on Azure AKS, reducing context retrieval overhead 3x across 500+ concurrent users, and cut LLM iteration cycles from 3 days to under 6 hours (12x).

Built a GPT-4/BERT recommendation engine with a RAG pipeline over 100,000+ educational resources, increasing course completion rates from 12% to 34% (183% lift). Architected LLM inference infrastructure on AWS EKS with model quantization for 8,000+ requests/day at 40% lower compute cost while maintaining 99.9% availability.

Fine-tuned BERT on political speech data using RLHF, achieving 89% accuracy and 25% bias reduction. Mentored 40+ graduate students and led hands-on AI/ML workshops to support continuous learning.

Redesigned a legacy investment platform into 15+ fault-tolerant Python microservices deployed via Docker, ECS, and Terraform on AWS, achieving 99.95% uptime and eliminating $1.2M in annual infrastructure costs. Built a FastAPI gateway with load balancing, Redis caching, and JWT auth to sustain 10,000+ req/s and sub-50ms p99 latency under peak load, and orchestrated Kubernetes migrations using HPA

Revamped 10+ web apps using React and TypeScript with lazy loading and code splitting, improving page load performance by 45%. Automated CI/CD with GitHub Actions, Docker, and AWS S3/CloudFront, reducing deployment time from 4 hours to under 30 minutes.

Pursued a Master of Science in Computer Science, covering coursework in program design, web development, data structures and algorithms, cloud, databases, NLP, and AI foundations.

Activities and societies: Winner/finalist at 4+ major technical hackathons; delivered full-stack AI prototypes within 48-hour constraints.

Earned a Bachelor of Technology in Information Technology, with experience building full-stack AI prototypes during multiple technical hackathons.