Skip to content

AI Agent Security · Full Stack · LLM Orchestration · ML

Mohib
Uddin.

Full-stack engineer building LLM agents and ML systems on Python + TypeScript.

Now
SWE @ Techstack Digital · since 2025
Based
Karachi · GMT+5 · remote-first
Rates
$35–55/hr · $40–60k FT
See the workRésumé · PDFmohibuddin9@gmail.com
§ 01

Working

Where I’ve been shipping.

  • Enhanced 3+ production CMS platforms (Payload CMS & Django CMS) with custom Next.js/TypeScript content blocks, reducing content management time by 40%

    Engineered IoT integration connecting 15+ Zigbee (nRF54) sensors for real-time data collection with 99.8% uptime

    Built cross-platform desktop app using Tauri (Rust) with React frontend for internal tooling

§ 02

Selected Work

Different problems, different tools.

01

◈AI / LLM Agents

ClawVet — Skill Security Scanner

6-pass static + LLM scanner for OpenClaw/Claude skill files. Catches obfuscated payloads regex misses by routing through a semantic Claude pass. v0.8.0 on npm, ~4.9k downloads to date.

Source
  • TypeScript
  • Node.js
  • Anthropic SDK
  • CLI
  • Static Analysis
  1. 02

    ◈AI / LLM Agents

    ferry — LLM migration reports

    CLI that diffs your prompts across providers and reports what breaks, so a team can switch LLM models without guessing. Published on npm.

    TypeScriptNode.jsAnthropic SDKOpenAI SDK
  2. 03

    ◈AI / LLM Agents

    FastAPI WebSocket Transcription

    Real-time speech-to-text backend for a Toronto client (ROSOFT). 16 kHz PCM streamed over a single WS, Whisper inference, partial transcripts pushed back on the same socket — bounded asyncio queues, structured close codes, clean disconnects under load.

    PythonFastAPIWebSocketsWhisper
  3. 04

    ◇Full Stack

    ERP & Inventory System

    Production CMS for accounts, inventory, suppliers, invoicing. Next.js + Payload CMS with role-based access and Docker deployment.

    Next.jsPayload CMSTypeScriptDocker
  4. 05

    ◌IoT & Embedded

    Smart Home Security System

    Multi-protocol IoT bridge (nRF54L15 + BLE 5 + Zigbee). Fingerprint auth latency cut from 3–4 s to 30–40 ms. Matter/Thread for ecosystem interop.

    nRF54L15BLE 5ZigbeeThread/Matter
    ~100×
§ 03

Stack

Tools I reach for — by job.

01Language

  • Python
  • TypeScript

02Backend

  • FastAPI
  • Django
  • Node.js
  • WebSockets
  • Celery

03Frontend

  • React
  • Next.js
  • Tailwind CSS

04AI / LLM Agents

  • Anthropic SDK
  • OpenAI SDK
  • LangGraph

05Security

  • Static Analysis
  • OWASP Agentic Top 10
  • SARIF

06AI / Vision & Speech

  • PyTorch
  • YOLOv5 / v8
  • RF-DETR
  • ByteTrack
  • OpenCV
  • Whisper

07Database

  • PostgreSQL
  • Redis

08IoT & Embedded

  • nRF54
  • ESP32
  • Raspberry Pi
  • Zigbee / BLE
  • Embedded Linux
  • FPGA (Vitis AI)

09DevOps

  • Docker
  • Git

Background

Education

BE in Computer & Information Systems Engineering

NED University of Engineering and Technology · Karachi

Published Research

Comparative Assessment of YOLO Nano Architectures for High-Speed Steel Detection

Archives of Science and Technology Research Journal (ASTRJ) — 97.8% accuracy

§ 04

Open Source

Merged into projects I use.

§ 05

Writing

Notes on what I'm building and watching.

  1. 01

    Featured · · 3 min read

    The Fable 5 ban should scare anyone shipping AI security tools

    Why one model getting banned is a warning for everyone building security tooling on top of LLMs.

    Read on Medium
  2. 02

    AI Is Already Passing Messages Humans Can't Read

    · 5 min ↗
  3. 03

    I Built a Real-Time Object Tracking Analytics Framework on Top of YOLO

    · 4 min ↗
  4. 04

    How to Train YOLOv13 on Custom Datasets

    · 5 min ↗
Aside

What a client said

Mohib successfully figured out what could be wrong and how to proceed in a project where others got stuck.
— Client via Upwork · AWS SES & Python Integration
§ 06

Contact

Hiring for full-stack, LLM, or ML work?
Send a brief.

Email

mohibuddin9@gmail.com

GitHub

@MohibShaikh

LinkedIn

/in/mohibuddin