Skip to content

Niaz Nafi Rahman

Product executive at US-Bangla Airlines. I build software that does more with less — and Bangla ambigrams that read two ways.

My seal. Turn it over and nothing changes — that is the whole idea.

Hover, tap, or press Enter.

Work

  1. Executive — Product

    US-Bangla Airlines Ltd · Dhaka

    September 2025 — present

    Product work for an airline: gathering requirements from the people who do the job, writing them up as specifications an engineering team can build from, prototyping the parts that need to be seen before they can be agreed, and running user acceptance testing. Day to day that means sitting between operations, UI/UX and engineering.

    Described generically on purpose. Specifics of the current work are the employer's to disclose, not mine.

    • Requirements
    • Specification
    • Prototyping
    • UAT
  2. Independent Ambigram Artist

    Ghurnilipi · Dhaka

    February 2020 — present

    An independent practice making original Bangla ambigrams, sold directly. I run the whole pipeline myself: taking the brief, drawing in Adobe Illustrator, pricing, revisions, and delivery.

    • Adobe Illustrator
    • Pricing
    • Client delivery
  3. Student Tutor

    BRAC University · Dhaka

    February 2024 — May 2025

    Taught 300+ students across lab sessions and one-to-one consultations in Python, object-oriented programming, and data structures and algorithms. 15 hours a week alongside a full course load.

    • Python
    • OOP
    • Data structures & algorithms

BRACU Alter at RoboCup Rescue 2024

A rescue rover that has to find hazard signs in a collapsed building and put them on a map, built by a university squad on a student budget. I led the five-person AI team inside that squad.

Read the case study

Ambigrams

Rotational ambigrams in Bangla script. One drawing, read two ways.

  • Two names in one drawing. Turn it over and the other person's name is the one you were looking at all along.

  • Twin spirals at opposite corners, so the composition is symmetrical under the half turn even before you read it.

  • সম্প্রীতিSompriti — harmony

    A word whose meaning is two sides agreeing, set as one drawing that reads two ways. The conjunct ম্প্র is three consonants stacked — it had to survive being turned over intact.

  • বাংলাBangla

    Two red discs, one above and one below, so the flag reads the same either way up.

  • শূন্যতাShunnota — emptiness

Why Bangla makes this hard

A rotational ambigram is one drawing that reads as one word the right way up and a different word turned 180°. Nothing changes but your point of view.

Bangla resists it in three specific ways. Conjuncts (যুক্তাক্ষর) fuse two or three consonants into a single shape, so the unit you are inverting is often not a letter at all. Matras hang above and below the line, and they do not survive a rotation unless the pairing was designed for it. And the মাত্রা — the headline stroke running across the top of the word — becomes a floor when you turn the drawing over, so it has to read as both.

Most letter pairs simply do not resolve. The work is finding the ones that do, and then drawing them so neither reading looks like the compromise.

The finished pieces are sold through my practice.

ghurnilipi.com

All 17 pieces

Projects

  • The art and the engineering meeting

    Bangla Ambigram Illustrator Panel

    An Adobe Illustrator panel that remembers letter pairings, for anyone drawing Bangla ambigrams — which in practice is me.

    Every ambigram is built out of letter pairings: shapes that read as one letter upright and a different one upside down. I was solving the same pairings over and over by hand. So the panel stores each pairing the first time it is drawn, finds it again by letter, and places it back on the artboard. Each finished word records which pairings built it. Everything lands on disk as readable files plus two spreadsheets, so the archive outlives the tool.

    • Adobe Illustrator UXP
    • Node.js
    • JavaScript
  • Co-authored undergraduate thesis · published on arXiv

    Bridging Human and Model Perspectives

    Political bias detection in news media using large language models — measured against what human annotators actually agreed on.

    A manually annotated corpus of news articles, used to compare how humans and four model families — GPT, BERT, RoBERTa and FLAN — judge political slant. Fine-tuned RoBERTa aligned closest with human labels among the transformer baselines; GPT agreed most strongly overall, zero-shot. The interesting result is the gap itself: humans and models disagree systematically about what counts as bias, not randomly.

    • Python
    • Hugging Face Transformers
    • LangChain
  • Full-stack EdTech · Nov — Dec 2024

    AI Tutoring Platform

    A study tool that generates questions from the material a student is actually reading, and marks them as they go.

    Contextual multiple-choice generation, a chatbot for follow-up questions, and feedback during a test rather than after it. Built on the Groq API because the latency budget for 'feels like a tutor' is small.

    • React
    • Express.js
    • Groq API

About

Niaz Nafi Rahman in a graduation gown, holding a mortarboard.

I read Computer Science at BRAC University and finished in 2025. I now work on product at US-Bangla Airlines, which mostly means turning what people actually need into something an engineering team can build.

The thread through everything else is the Bangla script. My thesis was about how language models read Bangla news; my side practice draws Bangla names as rotational ambigrams; and the tool I built for that practice exists because I kept re-solving the same letter-pair problem by hand. Same alphabet, three different problems.

Education

  • B.Sc. Computer Science

    BRAC University · July 2025 · CGPA 3.90 / 4.00

  • HSC

    Notre Dame College · July 2020 · GPA 5.00

  • SSC

    Ideal School and College · 2018 · GPA 5.00

Awards

  • Champion, BRACU Intra University Junior Programming Contest

    2022

  • Vice Chancellor's List and Dean's List, BRAC University

    2021 — 2025

Skills

Product
Requirements gathering · Stakeholder management · User stories · Wireframing · Agile · PRD documentation
Prototyping
Lovable · Gemini · React · Tailwind
Languages
Python (proficient) · Java · C++ · JavaScript · SQL
ML & vision
LangChain · Hugging Face Transformers · PyTorch · OpenCV · YOLO · NumPy · Pandas
Web & data
React · Express.js · Tailwind · HTML/CSS · MySQL · PostgreSQL
Robotics
ROS · Jetson Xavier · Raspberry Pi · RealSense · LiDAR
Creative
Design thinking · Adobe Illustrator · Bangla typography

Contact

Download CV