CS & Economics @ NYU building software at the intersection of AI systems, search, infrastructure, and applied ML.

I like working below the wrapper layer: retrieval systems, LLM infrastructure, storage engines, evaluation, and full-stack AI products.

AI workbench for pharmaceutical regulatory affairs Python Go TypeScript PostgreSQL pgvector LLMs

Built around two complementary workspaces:

• Research Studio — researches the public FDA record, combining cited Q&A, product-specific dossiers, regulatory monitoring, and evidence-backed research artifacts.
• Compliance Studio — a document-review workbench for analyzing internal CMC drafts against standards such as ICH, USP, 21 CFR, and internal SOPs, with findings anchored directly to source text.

The retrieval system scopes questions to the product under review, attaches passage-level provenance to FDA claims, and rejects unsupported factual claims instead of presenting them as grounded answers.

Git-style version control for any directory Java 21 Systems Storage

A dependency-free snapshotting system built from scratch with:

• content-addressed SHA-256 object storage
• immutable blobs and recursive tree objects
• parent-linked commit history
• snapshot-to-snapshot and working-directory diffs
• atomic object writes
• deduplication across files and snapshots
• streaming restore for large files
• symlink and executable-bit preservation
• integrity verification before destructive restores
• integration tests against real temporary filesystems

Single-GPU OpenAI-compatible LLM inference API Python vLLM FastAPI LLM Infrastructure

Built an OpenAI-compatible inference service around vLLM with:

• API authentication
• rate limiting
• FastAPI gateway
• GPU model serving
• OpenAI-compatible endpoints
• inference benchmarking
• cost-per-million-token analysis

Search and retrieval infrastructure Python Search Retrieval

Exploring the systems behind information retrieval and search infrastructure, with a focus on building retrieval rather than only consuming search APIs.

LLM Systems        ████████████████████
Search / Retrieval ████████████████████
Backend / Systems  ███████████████████░
Applied ML         ███████████████████░
AI Products        ████████████████████

I'm especially interested in:

LLM infrastructure · inference · retrieval · evaluation · backend systems · distributed systems · applied ML

💻 Portfolio 💼 LinkedIn 📫 rajahh7865@gmail.com