[ research · perception systems ]

Machine Learning Researcher · Computer Vision & Deep Learning

S. M. Navin
Nayer Anik

I research perception under uncertainty — how models keep seeing correctly when the signal is noisy, the labels are scarce, and the world shifts: forests read from drones, structure recovered from murky water, disease inferred from spectra.

Download CVRead the researchGet in touch
S. M. Navin Nayer Anik
subject: anik0.99
Open to research collaboration
[ Profile ]CS · BRAC University

Perception at the messy end

A machine learning researcher whose applied work at Cefalo grounds a research program on perception under uncertainty — noisy input, scarce labels, shifting distributions.

I'm a machine learning researcher working at the hard end of perception — turbid water, seasonal drift, scarce labels — where a model has to stay right as the world changes underneath it. My applied work at Cefalo (computer-vision and generative models for aerial imaging, underwater robotics, and pricing) is where these methods get pressure-tested at production scale.

That work sits alongside an active research record: a first-author paper on cross-domain fault diagnosis in Nature Scientific Reports, multi-label antimicrobial-resistance prediction from MALDI-TOF spectra under review, self-supervised vision transformers for anomalous sound detection in preparation, and an optimal-transport GAN thesis for medical image augmentation. Earlier, I led software for BRACU DUBURI, Bangladesh's first autonomous underwater vehicle.

Role
ML Researcher & Engineer · Cefalo
Based
Dhaka, Bangladesh · remote for Norway
Education
B.Sc. CSE, BRAC University · GPA 3.82
Thesis
Optimal-transport GAN for medical image augmentation
Focus
Perception under uncertainty · generative models · signal → structure

[ Research interests ]

Computer vision
Deep learning
Generative models
Optimal transport
Medical imaging
Transfer learning
Robotics

[ Recognition ]

  • RoboSub USA 2022

    Top 10 semifinalist · Outstanding Rookie

  • 4th IR Conference 2021

    Top 5 industrial projects

  • Dean's & VC's List

    5× recipient

[ Research ]04 papers

Publications

Peer-reviewed papers, manuscripts under review, and work in progress — across fault diagnosis, generative augmentation, and biomedical signals.

  1. 01
    Published
    Cross-Domain Intelligent Fault Diagnosis with Superlet Spectrograms and a Parameter-Efficient CNN-Transformer Hybrid

    S. M. Navin Nayer Anik, Md. Ehsanul Haque, Fahmid Al Farid, Mahe Zabin, Jia Uddin, Hezerul Bin Abdul Karim · Nature Scientific Reports, 2026.

  2. 02
    Published

    Optimal Transport Theory based GAN for Medical Image Augmentation and Classification

    S. M. Navin Nayer Anik · Thesis - BRAC University, 2023.

  3. 03
    Under review

    MSFormer: Multi-Label Antimicrobial Resistance Prediction from MALDI-TOF Spectra

    S. M. Navin Nayer Anik · Under review at PLOS One, 2026.

  4. 04
    In progress

    From Acoustic Signals to Visual Patterns: Advancing Anomalous Sound Detection Using Self-Supervised Vision Transformers

    S. M. Navin Nayer Anik · Manuscript in preparation, 2026.

[ Competitions ]

RoboSub International AUV Competition · 2022

Top 10 Semifinalist & Outstanding Rookie Team (USA)

Conference on 4th IR · 2021

Top 5 Industrial Projects

View on Google Scholar
[ Research projects ]04 projects

Research & robotics

Research systems from university and lab work — where the methods were built, tested, and taken to competition.

Underwater2019–2023

BRACU DUBURI — Autonomous Underwater Vehicle

Bangladesh's first autonomous underwater vehicle (AUV). Developed transfer-learning-based object detection for the rover using a Jetson Nano as the onboard processing unit. The team competed internationally, earning recognition as a top rookie team.

  • RoboSub USA 2022 — Top 10 Semifinalist & Outstanding Rookie Team
  • Real-time underwater object detection on edge hardware
  • 4th IR Conference 2021 — Top 5 Industrial Projects
PyTorch
Transfer Learning
Jetson Nano
OpenCV
ROS
Python
Generative2023

Optimal Transport Theory-based GAN

Built a Generative Adversarial Network grounded in Optimal Transport theory to synthesize realistic datasets, improving downstream classification model accuracy. Thesis project exploring the intersection of mathematical theory and generative deep learning for medical image augmentation.

  • Undergraduate thesis — BRAC University
  • Synthetic data generation for medical image classification
  • Bridging mathematical theory with practical deep learning
TensorFlow
GANs
Optimal Transport
Medical Imaging
Python
NumPy
Vision

Real-time Sleep Detection & Warning System

Computer vision system that monitors a driver's face in real-time and triggers an alert when signs of drowsiness are detected. Uses facial landmark detection and eye-aspect-ratio thresholds for accurate sleep state classification.

  • Real-time facial landmark tracking
  • Eye-aspect-ratio based drowsiness classification
  • Audio alarm trigger system
OpenCV
dlib
Python
NumPy
Signal

Sine-wave Predictor

Time series forecasting model that learns and predicts sine-wave patterns using a sequential neural network with Dense and SimpleRNN layers. A focused exploration of recurrent architectures for signal prediction.

  • Sequential model with Dense + SimpleRNN layers
  • Signal pattern learning and forecasting
TensorFlow
Scikit-learn
NumPy
Matplotlib
Python
[ Career ]2019 → now

Trajectory

From leading software on an autonomous underwater vehicle to shipping ML for international clients.

  1. Sep 2023 — Present

    Software Engineer I, AI/ML

    CefaloDhaka, Bangladesh (Remote for Norway-based clients)

    1. Trainee Software EngineerSep 2023 — Apr 2024
    2. Associate Software EngineerApr 2024 — Jun 2025
    3. Associate Software Engineer IIMay 2025 — Feb 2026
    4. Software Engineer IJan 2026 — Present

    Developing and deploying computer vision, generative AI, and optimization models for international clients in aerial imaging, underwater robotics, and dynamic pricing.

    • BioDrone — Developing CNN and transformer-based models for aerial forestry and agricultural image analysis, with data-centric quality analysis (3LC) and distributed training
    • Dobee — Built a generative AI story generation system with human-in-the-loop refinement, retrieval-and-synthesis pipeline, and conversational context injection
    • TicketCo Data — Formulated dynamic ticket pricing as constrained profit optimization; delivered rule-based PoC and designed ML forecasting roadmap
    • AquaRobotics — Developed underwater image classification with GAN-based synthetic augmentation to mitigate dataset bias and overfitting; reduced training cost by 70%
    • Built a full-stack blogging platform with JWT authentication, role-based access, and comprehensive unit testing (React, TypeScript, Express.js, Redux Toolkit)

  2. 2019 — 2023

    Sub-Team Lead

    BRACU DUBURIBRAC University, Dhaka

    Led the software sub-team for Bangladesh's first autonomous underwater vehicle (AUV).

    • Developed transfer-learning-based object detection for the underwater rover using Jetson Nano
    • RoboSub USA 2022 — Top 10 Semifinalist & Outstanding Rookie Team

  3. 2021 — 2023

    Machine Vision Engineer

    Anusondhani LabBRAC University, Dhaka

    Research and development in computer vision and deep learning for academic projects.

    • CNNs, transformers, and transfer learning for underwater and medical imagery
    • 4th IR Conference 2021 — Top 5 Industrial Projects

  4. 2020 — 2023

    Undergraduate Teaching Assistant

    BRAC UniversityDhaka, Bangladesh

    Assisted faculty in teaching computer science courses and mentoring students.

[ Applied work ]05 systems

Applied ML in production

Where the methods meet real constraints — production ML for international clients across aerial imaging, underwater robotics, generative AI, and pricing.

D·01Dec 2025 — Present

BioDrone

Active

Associate Software Engineer II, AI/ML

Developing CNN and transformer-based computer vision models for BioDrone's aerial imagery platform, enabling automated analysis of drone-captured forestry and agricultural imagery. Focused on model integration, domain generalization across seasonal/geographic variation, and scalable distributed training with data-centric quality analysis (3LC).

  • CNN and transformer models for forestry and agricultural aerial image analysis
  • Data-centric quality analysis (3LC) for large-scale aerial datasets
  • Scalable distributed training for high-resolution imagery experiments
PyTorch
CNN
Transformers
Computer Vision
3LC
Distributed Training
SageMaker
D·02Sep 2025 — Dec 2025

Dobee

Associate Software Engineer II, AI/ML

Built a generative AI system for interactive, context-aware story generation. Designed a retrieval-and-synthesis pipeline that transforms structured data into narrative summaries, with iterative human-in-the-loop refinement via conversational context injection for progressively more grounded outputs.

  • Generative AI narrative synthesis from structured data sources
  • Human-in-the-loop iterative refinement for controllable text generation
  • Context-aware story regeneration via conversational feedback
Vertex AI
LangChain
Generative AI
Prompt Engineering
Python
D·03Jul 2025 — Sep 2025

TicketCo Data

Associate Software Engineer II, AI/ML

Contributed to a proof-of-concept dynamic ticket pricing system, formulating profit maximization as a constrained optimization problem with configurable business rules. Delivered a real-time pricing recommendation engine and designed the research roadmap for ML-based demand forecasting to enable learned pricing policies.

  • Profit-maximization optimization with configurable business constraints
  • Real-time dynamic pricing recommendation engine (rule-based PoC)
  • Research roadmap for ML-based demand forecasting and learned pricing
Python
Time Series Forecasting
Optimization
Dynamic Pricing
D·04Feb 2024 — Jul 2025

AquaRobotics

Associate Software Engineer, AI/ML

Developed CNN-based underwater image classifiers for an autonomous sea-fish-farming rover, addressing overfitting and dataset bias in turbid, low-illumination aquatic imagery. Designed and implemented a GAN-based synthetic data augmentation pipeline to expand underrepresented classes and improve model generalization. Applied data-centric curation (3LC) to identify labeling errors and distribution skew. Scaled iterative model experimentation via distributed training, reducing compute cost by up to 70%.

  • GAN-based synthetic augmentation to mitigate underwater dataset bias and overfitting
  • Data-centric curation (3LC) for label quality and distribution analysis
  • Reduced training compute cost by up to 70% via distributed spot-instance training
PyTorch
CNN
GANs
Computer Vision
3LC
Distributed Training
SageMaker
D·05Sep 2023 — Feb 2024

Blogging Website (Full-Stack)

Junior Software Developer

Built a full-stack blogging platform with JWT authentication, role-based CRUD operations, and comprehensive unit test coverage across frontend and backend.

  • JWT-based secure authentication
  • Full CRUD operations with role-based access
  • Comprehensive unit test coverage
Express.js
Sequelize
React
TypeScript
Redux Toolkit
Tailwind CSS
React Router
Jest
[ Methods & tools ]6 classes

Capability matrix

The instruments — languages, frameworks, and infrastructure behind the research and applied work.

▸Languages
Python
JavaScript
TypeScript
Java
Bash
▸ML / DL
PyTorch
TensorFlow
Keras
Scikit-learn
OpenCV
GANs
Transformers
Computer Vision
Transfer Learning
▸Cloud / MLOps
3LC
Distributed Training
SageMaker
CI/CD
Spot Instances
AWS
GCP
Docker
Lambda
EventBridge
CloudWatch
Google Colab
▸Data
Pandas
NumPy
Matplotlib
Seaborn
Apache Spark
▸Databases
SQL
MySQL
MongoDB
Firebase
BigQuery
▸Web / Tools
Node.js
Express.js
React.js
Redux Toolkit
Tailwind CSS
REST APIs
Jest
Git
Bitbucket
ROS
[ Contact ]● open to research collaboration

Let's talk

Research collaboration, a hard perception problem, or just to compare notes — send a note or reach me on any channel below.

[ Channels ]

[ Send a message ]