FF FAHIM FAISAL
--:--:-- RUN

05 — About

One question, asked
in three mediums.

I’m Fahim — an autonomy engineer from Dhaka. I started where machines are hardest to hide: something that has to leave the ground. Rockets taught me guidance and control; rovers taught me manipulation and vision; underwater vehicles taught me to trust nothing I hadn’t measured. The thread through all of it is the same question — how does a machine understand the world well enough to act in it?

Today I build the whole stack: the firmware on the board, the estimator that fuses cheap sensors, and the neural net that sees. I care about the seam between software and physics — the place where an elegant idea meets a vehicle that either moves or doesn’t.

Fahim Faisal

Muhammad Fahim Faisal

Autonomy & robotics engineer

  • LOCATIONDhaka, Bangladesh
  • FOCUSAutonomy · robotics · ML
  • STATUS● Open to autonomy / robotics / ML roles
the field // mongla e-stop rec ●
Hands on the vehicle during an emergency-stop, mid-run
Not a render. Mid-run, hands on the machine, the abort path proving it works. fh1m

// experience

  • BRACU Duburi

    Team Lead — Engineering · former Sub-Team Lead, AI & Machine Vision

    Oct 2023 — Sep 2026 Dhaka
    • Rebuilt the vision stack on state-of-the-art CNN architecture with a custom dataset pipeline.
    • Created a training pipeline that streamlines custom models with advanced hyperparameter control.
    • Integrated Domain-Specific Pretraining and Source-Free Domain Adaptation.
    • Deployed on edge devices for efficient on-vehicle inference.
    • Built in-house dev tooling for management and deployment.

  • BRACU Mongol-Tori

    Contributor — Autonomous Subteam

    Dec 2024 — May 2025 Dhaka
    • Implemented inverse kinematics for a 2-DOF arm for autonomous typing tasks.
    • Built and deployed a lightweight vision unit for arm alignment, separate from the rover.
    • Wrote custom OCR + image processing to identify and differentiate keys.
    • Built an in-house GUI and sim environment for calibration and early testing.
    • Managed the dataset pipeline for URC 2025.

  • Amateur Experimental Rocketry (AERD)

    Core Member — A&GNC, Trajectory Prediction, Parachute & Procurement

    Feb 2023 — Feb 2025 Dhaka
    • Implemented a ROS 2 (Humble) stack for data acquisition.
    • Worked with flight controllers (Pixhawk, Betaflight); led V1 of an in-house flight computer for a monocopter UAV.
    • Worked on a VSLAM software stack for GPS-denied autonomous navigation.
    • Ran trajectory prediction with RocketPy against real flight-test data.

  • BRACU Rocketry — LASSET Lab

    Sub-Team Member — Controls & Trajectory Prediction

    Sep 2024 — Feb 2025 BRAC University
    • Built an in-house PID controller for a TVC gimbal; submitted the demo to the Research Excellence Awards Initiative for funding.
    • Implemented a 2-axis controller.
    • Built a custom GUI for tracking and monitoring.

  • Dubotech

    Jr Autonomous Engineer

    Oct 2025 — Feb 2026 Dhaka
    • Developed custom firmware for sonar tech.
    • Created a navigation stack for deep-water monitoring.
    • Built and maintained backends for the sonar stack; started in-house tooling documentation.

  • Thinkit Inc

    Web Dev Intern

    Mar 2021 — Jun 2021 Dhaka
    • Wrote back-end code for a transition to MongoDB.
    • Fixed front-end bugs.

  • ICT Club — SOS Hermann Gmeiner College

    General Secretary

    Jan 2022 — Sep 2022 Dhaka
    • Ran the club end to end — seminars, programs, operations.
    • Automated content management with the YouTube / Facebook APIs.

// education

BRAC University

BSc Computer Science — Minor: Electrical & Electronic Engineering

Sep 2023 — present Dhaka, Bangladesh

Currently in the 6th semester.

IU International University of Applied Sciences

BSc — Autonomous Systems

2024 Bad Honnef, Germany

One semester, on campus.

intermission a fun fact, grounded in real engineering
influences / shoulders who taught me to build real work · real debt

ten people, one reflex each

George Hotz
b. 1989

George Hotz

geohot · tinygrad, comma.ai
“Complex things eventually collapse under their own weight.”
the work tinygrad — “the RISC of the ML stack” — and openpilot, self-driving on a phone.
what it changed in me The reason the stack stays small. Every new layer has to justify itself against a delete. I attempted his comma.ai calibration challenge for the same reason.
their writing — geohot.github.io/blog ↗ the source ↗
published / indexed peer-reviewed & indexed

Research, with my name on it.

OpenReview 2025

Color is Also Class Evidence: Re-evaluating Domain Generalization for Underwater Object Detection

T. M. Wasif Ishmam, Yaad Kamrul Bari, Muhammad Fahim Faisal, Md Khalilur Rhaman

Advanced vision systems for AUVs — tested against the RoboSub 2025 / 2026 footage.

computer-visiondomain-generalizationunderwaterobject-detection
READ THE PAPER ↗
Universe International Journal of Interdisciplinary Research, Vol. 4 Issue 11 2024

E-Commerce Recommendation System Based on Demand Clues

I. N. Reeve, T. Alam, M. Rahman Munia, A. Atif Showmik, M. F. Faisal

Recommendation systems to improve customer experience, with an international group of researchers.

recommender-systemse-commerceml
READ THE PAPER ↗

A short list of things I believe, and keep testing against the machine.

  1. 01 I like machines that have to deal with reality.
  2. 02 I like understanding what sits beneath the abstraction.
  3. 03 I like taking expensive or opaque systems apart and rebuilding them from first principles.
  4. 04 I like robotics because there is nowhere to hide: eventually the machine has to move.
  5. 05 I like software because it lets an idea become behavior.
  6. 06 I like hardware because physics gets the final vote.
  7. 07 I like open source because knowledge should compound.
  8. 08 I like difficult problems because they expose what I don’t understand.
  9. 09 And I am still building.