Lecturer / Researcher / Software Engineer

Algorithms, Learning Systems, and Academic Technology.

My work connects theoretical computer science, data-centric AI, production software, academic quality systems, and research-driven teaching.

View Research Outputs Google Scholar Download CV

16: Scholar entries
31: GitHub repositories

Academic Portfolio

A research-centered profile with software-building depth.

Tanzeem Rahat is a Lecturer in the Department of Computer Science at American International University-Bangladesh. This portfolio presents his work as a researcher-builder: one who studies algorithms and AI, builds software systems, supports academic quality practice, and turns research ideas into working tools.

Research Themes

Four lenses for the work.

Algorithmic Research

Sequence algorithms, subsequence structure, output-sensitive dynamic programming, pattern matching, and online assignment problems.

Learning Systems

Computer vision, multimodal machine learning, Bengali NLP, gesture recognition, road-safety AI, medical imaging, and reproducible evaluation.

Datasets and Benchmarks

Smartphone RGB gesture data, metadata-driven evaluation, mobile camera constraints, MWPD-style road imagery, and benchmark design for real-world AI.

Academic Technology

Teaching systems, OBE workflows, IQAC assessment, programming quizzes, and academic analytics for evidence-based improvement.

Research Outputs

Publications, preprints, and datasets in one academic record.

IWOCA 2025 / LNCS 15885

A Space-Efficient Algorithm for Longest Common Almost Increasing Subsequence

Space-efficient exact algorithms for LCaIS, including an O(nmL)-time, O(n + mL)-space result and a match-sensitive faster regime.

Quick brief

Problem: Find common subsequences that are almost increasing.
Core idea: Exploit structural constraints and sparse matches.
Result: Space-efficient exact algorithms for LCaIS.
Open direction: Faster bounds under stronger sparsity assumptions.

Open paper journey

arXiv / revised Jan 2026

The Longest Common Bitonic Subsequence

Match-sensitive dynamic programming for LCBS, moving from dense grid thinking toward output-sensitive structure.

Quick brief

Problem: Compute longest common bitonic subsequences.
Core idea: Use match-sensitive structure instead of dense tables.
Result: Output-sensitive algorithmic framing for LCBS.
Open direction: Sharper subquadratic regimes for broader inputs.

Open paper journey

arXiv / Jan 2026

Permutation Matching Under Parikh Budgets

Linear-time detection, packing, and disjoint selection for jumbled pattern matching and Parikh-vector budget problems.

Quick brief

Problem: Detect and select matches under symbol-count budgets.
Core idea: Model matching as Parikh-vector resource allocation.
Result: Linear-time detection, packing, and disjoint selection.
Open direction: Apply the budget view to nearby string problems.

Open paper journey

arXiv / Feb 2026

Online Facility Assignment on Discrete Grid Graphs

Competitive analysis of capacity-aware online assignment on grid graphs, including failure modes and batching remedies.

Quick brief

Problem: Assign online arrivals to capacitated facilities on grids.
Core idea: Analyze how geometry and depletion shape decisions.
Result: Performance bounds and remediation strategies.
Open direction: Sharper guarantees for semi-online batching.

Open paper journey

arXiv / 2026

Expected Cost of Greedy Online Facility Assignment on Regular Polygons

Expected-cost analysis for greedy online facility assignment on regular polygons using recurrence, symmetry, dynamic programming, and simulation.

Quick brief

Problem: Measure greedy assignment cost on regular polygons.
Core idea: Use symmetry, recurrence, and dynamic programming.
Result: Exact expected-cost analysis with simulation support.
Open direction: Scale symbolic costs to larger polygon families.

Open paper journey

Research Square / 2024

Systematic Literature Review on Bengali Grammar Correction

A review of Bengali grammar correction systems, methods, evaluation metrics, language-specific challenges, and future research directions.

Quick brief

Problem: Clarify the Bengali grammar correction research landscape.
Core idea: Organize methods, datasets, metrics, and limitations.
Result: Systematic review for low-resource grammar correction.
Open direction: Comparable benchmarks for Bengali correction systems.

Open review journey

Systems and Soft Computing / 2026

HCF-Net

Hierarchical and conflict-aware fusion for robust multimodal sentiment understanding, published in Systems and Soft Computing.

Quick brief

Problem: Fuse multimodal sentiment signals despite conflicts.
Core idea: Use hierarchy and conflict awareness during fusion.
Result: Robust multimodal sentiment understanding framework.
Open direction: Test conflict-aware fusion across noisy real-world media.

Open paper journey

IEEE / 2026

Cl-CNN for Ethnic Language Recognition

A low-resource character-level CNN for Bangladeshi ethnic language recognition in scarce-data language settings.

Quick brief

Problem: Recognize Bangladeshi ethnic languages with limited data.
Core idea: Model language signals at the character level.
Result: Low-resource CNN approach for ethnic language recognition.
Open direction: Expand benchmarks for underrepresented local languages.

Open paper journey

Results in Engineering / 2025

Robust Multi-Weather Pothole Detection

An enhanced YOLOv9 model trained on the MWPD dataset for pothole detection under varied weather and road conditions.

Quick brief

Problem: Detect potholes reliably across weather conditions.
Core idea: Train enhanced YOLOv9 on multi-weather road imagery.
Result: Robust pothole detection for road-safety AI.
Open direction: Test transfer across regions and road textures.

Open paper journey

ICCIT 2025

AutoContDamage

Automated container damage detection in ports using an RF-DETR model for enhanced operational efficiency.

Quick brief

Problem: Detect container damage in busy port settings.
Core idea: Apply RF-DETR to operational damage-detection imagery.
Result: Automated visual inspection pipeline for ports.
Open direction: Handle subtle defects under clutter and occlusion.

Open paper journey

NSysS 2025

ChainGraph-DL

Data-driven causal inference for graph structure learning, accepted at the 12th International Conference on Next Generation Computing, Communication, Systems and Security.

Quick brief

Problem: Learn graph structure with causal signals.
Core idea: Use data-driven inference for graph learning.
Result: ChainGraph-DL for causal graph structure learning.
Open direction: Study robustness under noisy observational data.

Open paper journey

IEEE / 2026

Autoencoder-ViT Anomaly Detection

Combining autoencoders with vision transformers for anomaly detection in complex industrial images.

Quick brief

Problem: Detect anomalies in visually complex industrial images.
Core idea: Combine reconstruction learning with ViT representations.
Result: Hybrid industrial image anomaly-detection pipeline.
Open direction: Improve defect localization under domain shift.

Open paper journey

IEEE / 2026

FedEPL+

Carbon-aware client selection with valid differential privacy for federated learning.

Quick brief

Problem: Select federated clients while balancing privacy and carbon cost.
Core idea: Make client selection carbon-aware and privacy-valid.
Result: Federated learning selection strategy with DP constraints.
Open direction: Study fairness across clients with different energy profiles.

Open paper journey

IEEE BECITHCON 2025

Risk-Calibrated YOLO

A lightweight pipeline for reliable pneumonia detection in chest X-rays using risk-calibrated YOLO-style medical imaging.

Quick brief

Problem: Detect pneumonia while exposing model risk.
Core idea: Pair lightweight YOLO detection with risk calibration.
Result: Reliable chest X-ray pipeline for medical AI.
Open direction: Communicate uncertainty for clinical review workflows.

Open paper journey

IEEE Data Descriptions / Scholar indexed

RGB Video of Natural Hand Gestures Dataset (RGB-NHG)

A data descriptor and CC BY 4.0 smartphone RGB dataset with 312 clips across 13 natural hand gestures, lighting conditions, and capture views.

Quick brief

Problem: Create a citable RGB gesture dataset for mobile AI research.
Core idea: Capture natural gestures with smartphone constraints.
Result: Descriptor entry plus 312 clips across 13 gesture classes.
Open direction: Evaluate fusion and low-data recognition settings.

Open dataset journey

Research Map

How the research threads connect.

Algorithms

AI and Data

Academic Technology

Academic Profile

Research, teaching, and software engineering in one profile.

The portfolio is intentionally academic, but not only a list of papers. It shows a Lecturer and researcher who also builds software systems, datasets, pipelines, and academic tools.

Lecturer and Researcher

Faculty role at AIUB with research across algorithms, learning systems, datasets, and academic technology.

Software Engineer

Experience building production web, native, monitoring, scheduling, CRM, and enterprise systems before and alongside academic work.

Research Software Builder

Uses software engineering skill to support reproducible experiments, data organization, dashboards, teaching systems, and deployable AI workflows.

Software Skills

Engineering experience behind the academic work.

Languages

Comfortable across research scripting, systems work, web development, and application engineering.

Frontend and Native Apps

Worked on web, dashboard, and mobile application projects across modern JavaScript and native stacks.

Backend, Data, and Cloud

Production systems, APIs, databases, deployment workflows, and infrastructure tools used across professional and research projects.

Research and Academic Tooling

Algorithm prototypes, dataset organization, benchmark scripts, OBE workflows, programming assessment, and dashboards for academic quality work.

Production Projects

Built or led systems including Network Monitoring System, Generator Maintainer, SmartThink, Bit-Q, Dutch-Courage, ChimeTime, and Vromon.

View LinkedIn

Open Development

GitHub profile shows 31 repositories and practical work across JavaScript, C#, PHP, React practice apps, and web application experiments.

View GitHub

Teaching and Academic Quality

Computer science education with research-grade tooling.

Teaching, assessment, and quality assurance are part of the same systems mindset: create evidence, make progress measurable, and help students and programs improve.

Lecturer, Department of Computer Science

Faculty role at American International University-Bangladesh, linked with algorithms, software engineering, and intelligent systems.

Program Assessor, IQAC

Joined as Program Assessor of IQAC, AIUB on June 1, 2026, supporting program assessment and quality assurance practice.

Teaching Automation

Programming quizzes, OBE workflows, student analytics, and academic dashboards belong naturally beside the research agenda.

Open Problems

Questions that make the research agenda visible.

LCaIS

Can LCaIS be made faster when matching and compatible pairs are sparse?

A positive result would make the theory paper easier to connect to large structured sequence data.

LCBS

Can bitonic subsequence algorithms be pushed below dense quadratic behavior on broader input classes?

This is the natural pressure point between exact dynamic programming and practical scalability.

Parikh Budgets

What other string problems can be reframed as exact resource-budget packing?

The budget view may unify several matching, packing, and selection problems under one toolkit.

Online Facility Assignment

Can semi-online batching yield sharp competitive guarantees on discrete grids?

Batching is a realistic intervention when fully online greedy decisions become unstable.

Polygon OFA

Can exact expected-cost recurrences scale beyond small regular polygons without losing interpretability?

Symbolic models are useful only if they stay explainable as the geometry grows.

Decision Models

When do capacity-aware heuristics fail because local geometry collapses?

These failure cases can guide remediation strategies before a system is deployed.

Gesture Recognition

Can full-body and hand-only streams improve low-data gesture recognition?

Late fusion could turn a small RGB dataset into a stronger multimodal benchmark.

Bengali NLP

Which benchmark designs make Bengali grammar correction systems comparable?

Better benchmarks would help future systems compete on evidence, not only examples.

Multimodal Sentiment

How should fusion models decide when text, audio, and visual signals disagree?

Conflict-aware fusion is strongest when it can explain which modality it trusted and why.

Ethnic Language Recognition

Can character-level models scale to more Bangladeshi ethnic languages with only small corpora?

Low-resource recognition needs benchmarks that respect local scripts, spelling variation, and limited labeled text.

Road-Safety AI

How well do pothole detectors generalize across weather, geography, and road-surface texture?

Generalization is the difference between a model demo and a road-safety tool.

Port Automation

Can RF-DETR-style detectors handle subtle container damage under cluttered seaport conditions?

This turns object detection into an operational reliability problem.

Industrial Anomaly Detection

Can autoencoder and vision-transformer features localize rare defects in complex industrial scenes?

The useful system must find subtle anomalies and still explain where the visual evidence sits.

Federated Learning

Can client selection reduce carbon cost without weakening privacy guarantees or model fairness?

Sustainable federated learning has to balance energy, privacy, participation, and learning quality together.

Medical Imaging

How should lightweight pneumonia detectors expose calibrated risk rather than only class labels?

Risk communication is essential when AI outputs enter clinical review.

Academic Technology

Can assessment analytics reveal OBE gaps early enough to guide teaching decisions?

Earlier signals can make quality assurance useful during a course, not only after it.

Contact

For research collaboration, datasets, software systems, or academic technology.

This portfolio is open to collaboration across algorithms, learning systems, datasets, teaching tools, and research-grade software.

tanzeemrahat80@gmail.com Google Scholar LinkedIn GitHub AIUB Profile Download CV