Mujahid.

Quantitative research · Embedded & real-time systems

Signals that
clear the
threshold.

I build systems that have to measurably work — trading signals that pass hard statistical gates, and firmware that meets deadlines it cannot miss.

Role
Quantitative research consultant
& embedded systems engineer
Focus
Alpha research · RTOS & firmware
Edge AI · applied cryptography
Status
Open to freelance & full-time
Contact
mj@mujahidr.com

What I work on

Three domains, one habit: measure it

Quantitative research

Finding statistical edge in market and alternative data, then proving it survives out-of-sample testing, correlation screening, and hard submission thresholds — not just backtest well.

Python · NumPy · SQLite
Genetic search · TF-IDF retrieval
Signal & correlation analysis

Embedded & real-time

Firmware and multi-processor architectures where timing is a correctness requirement. Splitting work between an application processor and a microcontroller so nothing blocks what has to happen now.

STM32 · Zephyr RTOS
Linux MPU · Arduino
Sensors · actuation · HMI

Systems & applications

The full stack behind a working product: servers, encryption, native clients and the deployment around them. Built to ship to real users, not to sit in a demo folder.

Linux · Python servers
Kotlin · Jetpack Compose
End-to-end encryption

Selected work

Four projects, four different constraints

Alpha research pipeline

WorldQuant BRAIN · research consultant

Built brainminer, a Python pipeline that generates candidate trading signals, simulates them at scale, and screens them against the platform's live submission gates before a human ever looks at one. Genetic search runs across separate data-family populations; a TF-IDF retrieval layer maps a research hypothesis onto candidates from a field catalogue of roughly 90,000 fields. A local correlation stage decides which surviving signals are independent enough to be submitted together. Signals produced by this pipeline have been accepted into production.

  • Python + SQLite
  • Parallel simulation workers
  • Genetic + retrieval search
  • Live gate screening

CarromSight

Edge-AI referee & coach

A retrofit module that turns a traditional wooden carrom board into a self-refereeing, coaching board — no new table required. A dual-processor architecture splits the problem: a Linux MPU runs the vision stack, tracking fast-moving coins and hand landmarks, synchronised with an STM32 running Zephyr RTOS so physical response carries no perceptible latency. Players aim with 3D hand gestures mapped onto the board plane, with no wearables; the board answers with LED aim projection and buzzer foul alerts.

  • Linux MPU + STM32
  • Zephyr RTOS
  • MediaPipe hand tracking
  • LED + audio feedback

Latch

Encrypted messenger

An end-to-end encrypted messaging app where your identity is a key pair held on your device, not a phone number — so there is no directory to leak and no number to hand over. Built in stages: server and crypto core, browser client, device-transfer and safety-number flows, hardening, then packaging. A native Android client in Kotlin and Jetpack Compose now runs against the same server.

  • Python server
  • Device-key identity
  • Safety-number verification
  • Kotlin · Jetpack Compose

UNO Q console

Bidirectional device messaging

A two-way text and media console for the Arduino UNO Q: post messages and photos to the board from a phone or laptop, and read what the board has back on either. A small project, but the useful kind — it closes the gap between a board on a desk and the device already in your hand.

  • Arduino UNO Q
  • Linux on device
  • Web client

Available for work

Have something that has to actually work?

mj@mujahidr.com

Best for quantitative research, embedded and real-time firmware, and full-stack builds around either. Tell me the constraint you're up against and I'll tell you honestly whether I'm the right person for it.