Portfolio  /  LinkedIn  /  Email  /  Writing  /  ORCID

_{Domains  ·  Selected work  ·  Research  ·  Experience  ·  Toolkit  ·  Contact}

I'm Chinmay Shringi, a full-stack engineer and researcher in New York. I build across AI agents, secure systems and developer tools: runtimes that route work to the right hardware, agent workflows that can be reviewed and rolled back, and small tools that remove friction from everyday engineering.

Explore Systems Atlas, an interactive map of projects, research and shared ideas. Search the index, inspect an architecture, or explore the illustrative GRAFT routing lab.

• 01 / Systems. Distributed inference runtimes, WebAssembly sandboxing in Rust and C, and microservice back ends.
• 02 / Intelligence. LLM agent orchestration, retrieval-augmented generation over vector and graph stores, and small-model collaboration.
• 03 / Security. Red-team agent workflows, Dockerized cyber ranges, and defensive AI evaluation on live hosts.

Field entries, each with a public link.

SourceSnap  _{BROWSER EXTENSION TYPESCRIPT}
Alt+Click any element in a running app to open its source file in your editor. Works with React, Vue, Svelte, SolidJS and Preact, with no bundler plugin required.

Inspect SourceSnap in the Atlas.

Shoemaker Elves  _{AGENT ORCHESTRATOR PYTHON}
Breaks large projects into atomic tasks. An LLM plans a batch, a coding agent executes the tasks in sequence, and the LLM reviews results before planning the next batch.

Inspect Shoemaker Elves in the Atlas.

SLM Swarm  _{RESEARCH PLATFORM DOCKER FASTAPI OLLAMA}
A reproducible setup for comparing swarms of small language models against a single large model on summarization, using embedding-based consensus and standard metrics.

Persona AI  _{RAG PYTHON}
A self-hosted Gmail assistant: semantic search and question answering over your inbox with hybrid retrieval and a pluggable model backend.

Hybrid RAG  _{PGVECTOR NEO4J NODE.JS}
PDF question answering that combines pgvector semantic search with Neo4j knowledge-graph retrieval.

Spotlight Security Benchmark  _{AI EVALUATION SECURITY}
Benchmark work on evaluating AI agents at defensive security tasks on live hosts. I am listed among the authors.

_{More on the repositories tab.}

GRAFT: gRPC-Routed Agent Framework for Tasking in Edge and Personal Devices
_{C. Shringi, A. Hillel-Tuch, S. Rizwan  ·  ACM CAIS 2026  ·  DOI  ·  PDF}
A runtime for agent tasks on edge and personal devices. It routes work by capability across CPU, GPU and NPU targets over sessioned gRPC and keeps execution inside safety bounds.

Branch-Commit-Validate: A Git-Inspired Workflow for Autonomous Red-Team Agents
_{Springer CCIS 2937, proceedings of CSCE 2025 (2026)  ·  DOI}
Proposes a branch, commit and validate workflow so red-team agent actions can be reviewed, compared and rolled back like code.

An Advanced AI-Driven Database System
_{EDULEARN25 (2025)  ·  IATED library  ·  DOI 10.21125/edulearn.2025.1294}

M.S., New York University  _{2024 to 2026}

My experience spans research and product engineering: Rust, C and WebAssembly isolation work on the Lind project, Dockerized cyber ranges for LLM security evaluation, Node.js and Spring Boot microservices, WebRTC, and distributed consensus research.

Working on agents, secure systems or developer tooling? I'm glad to compare notes, review an idea or collaborate. Reach me at cs7810@nyu.edu, on LinkedIn, or through my portfolio.