Skip to content
View mansoor-mamnoon's full-sized avatar

Block or report mansoor-mamnoon

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
mansoor-mamnoon/README.md

Mansoor Mamnoon

SWE Intern @ Databricks · UC Berkeley EECS Honors, Class of 2027

Software engineer focused on distributed infrastructure, networking, low-latency C++, and ML security tooling.

Portfolio · Resume · LinkedIn · Email


Background

  • Databricks Traffic Platform — SWE intern focused on networking, security, and performance infrastructure in Go/Rust.
  • Amazon SDE Intern — shipped real-time IoT infrastructure processing 10K+ events/min across 100+ sensors with sub-second dashboard latency.
  • UC Berkeley EECS Honors — CS, GPA 3.973; coursework in OS, compilers, databases, algorithms, ML, and computer architecture.
  • CS 61C Senior Mentor & Content Lead — wrote worksheets and taught RISC-V, C, caching, pipelining, and memory hierarchy.

Featured Systems

MCP security proxy enforcing taint-aware least privilege for tool-using LLM agents.

  • Treats prompt injection as a control-flow and policy-enforcement problem across tool calls.
  • Seven-layer defense pipeline: capability gating, taint policy, secret-flow guard, argument scanning/sanitization, and output injection checks.
  • 4,200+ local eval cases; attack success reduced from 100% to 16% in baseline passthrough setting.

Stack: Python · FastAPI · MCP Protocol · NumPy · pytest


Exchange-style matching engine and quantitative replay pipeline.

  • Price-time priority matching with FIFO fairness, deterministic replay, snapshot/resume recovery, and strategy tooling (VWAP, TWAP, POV, Iceberg).
  • Performance work: slab allocators, branch reduction, cache-hot layouts, CPU pinning.
  • 20M+ msgs/sec, p50 ≈ 0.04µs (synthetic benchmark; hardware spec and harness in README).

Stack: C++20 · Python · Docker · Streamlit · GitHub Actions


Zero-config desktop IDE for deploying JS/TS APIs across Cloudflare, AWS, and Vercel.

  • Monaco editor, generated infrastructure-as-code, real-time deploy logs, deployment history, and provider-specific environment configuration.
  • Built to remove dashboard/CLI context switching from multi-cloud serverless deployment.

Stack: TypeScript · Electron · React · Pulumi · Monaco Editor · AWS · Cloudflare · Vercel


Technical Focus

Systems / Low-latency       C++20 · C · memory allocators · cache optimization · replay systems
Distributed Infrastructure  Go · Rust · networking · control planes · performance
Backend & Cloud             Python · TypeScript · AWS · Docker · Pulumi · Terraform
ML / LLM Security           MCP · prompt injection defense · taint-aware policies · agent tooling

Open to 2027 new-grad roles in systems/infrastructure, backend/platform, networking, low-latency, and ML security infrastructure.

mansoormmamnoon@berkeley.edu · mansoor-mamnoon.github.io · linkedin.com/in/mansoormamnoon

Pinned Loading

  1. limit-order-book limit-order-book Public

    High-performance limit order book engine with C++ core and Python SDK. Processes 20M+ msgs/sec with µs latency. Supports real crypto/equity data replay, spread/imbalance/impact analytics, and backt…

    C++ 62 21

  2. codeglance codeglance Public

    The 10-second codebase tour. Understand any repo in one command.

    TypeScript 64 4

  3. LLMFirewall LLMFirewall Public

    Defense-in-depth against prompt injection in tool-using LLM agents. Semantic drift detection catches synonym-evading attacks that bypass regex. Self-play red-team, calibrated scoring, ablation study.

    Python

  4. edge-deployer edge-deployer Public

    Desktop IDE to write and deploy serverless APIs to Cloudflare Workers, AWS Lambda@Edge, and Vercel Functions. Built with Electron, React, and TypeScript.

    TypeScript 2 1

  5. offline-rl-agent offline-rl-agent Public

    Offline RL benchmark project featuring a custom Gym environment, dual observation modes, reward shaping, and real-time PyGame rendering

    Python 1