Aeyan Ashraf

Hi, I’m Aeyan Ashraf—a machine learning researcher and M.S. candidate in Computer Science at the University of Massachusetts Amherst. I build grounded, trustworthy LLM systems that blend retrieval, evaluation, and deployment rigor, with recent work spanning financial question answering, adversarial ML, and medical imaging.

At Goldman Sachs, I designed a production-scale retrieval-augmented generation (RAG) pipeline for SEC 10-K filings, pairing LangChain automations with GPT OSS 120B and LLM-as-a-judge quality gates. Earlier, I explored backdoor defenses at Binghamton University and co-created TeliNet 2.0 for large-scale radiology triage at the University of Maryland.

What excites me most is closing the loop between research insights and usable ML infrastructure—standing up ingestion jobs, benchmarking pipelines, and evaluation harnesses that make reliability measurable.

Current focus

• Retrieval + Generation: multi-hop QA, grounded answer tracing, RAG benchmarking.
• Model Safety: adversarial robustness, LLM evaluation, automated quality control.
• ML Platforms: data orchestration, scalable inference APIs, MLOps instrumentation.

Always open to chat about

• Collaborations on interpretable or trustworthy ML systems.
• Opportunities to bring cutting-edge research into production workflows.
• Speaking, mentoring, or open-source projects that help democratize ML education.

If that resonates, feel free to reach out—email is quickest. Let’s build something impactful.