Portrait of Jabin Tasnim Urmy
26.5× speedup · OpenMP
CUDA · MPI · PyTorch
SnT · PCOG

Hello, I'm

Jabin Tasnim Urmy

I build scalable, intelligent software that connects high-performance computing, machine learning and cloud-native systems.

Hire Me View Projects
0+Years Experience
0+Major Projects
0+Technologies
0×Best Speedup Delivered
scroll

What I do

Services

High Performance Computing

Designing and optimising parallel applications using MPI, CUDA and OpenMP. Building scalable HPC pipelines on clusters and cloud platforms with thorough performance benchmarking and tuning.

AI / Machine Learning

Developing end-to-end AI solutions with PyTorch and TensorFlow, including deep learning architectures for computer vision and natural language processing, and deploying models for real-world inference.

Software Development & DevOps

Building robust applications and APIs with Python, C++ and Java. Automating workflows with CI/CD pipelines, Docker and Kubernetes to ensure reliable, reproducible deployments.

Data Analytics & Business Intelligence

Transforming raw data into actionable insights using Python, SQL, Power BI, Tableau, and Excel. Experienced in data cleaning, KPI reporting, dashboard development, ETL workflows, and large-scale analytics with Spark.

Web Development

Creating responsive full-stack web applications using Node.js, React and modern web technologies.

Who I am

About Me

Profile photo of Jabin Tasnim Urmy
Based in Luxembourg

I'm a software engineer completing a Master's in High-Performance Computing at the University of Luxembourg, working at the intersection of HPC, AI/ML, cloud, and data systems. I turn complex technical problems into software that performs reliably in practice.

At SnT, I develop and optimise C++ and Python software, focusing on parallel and GPU-accelerated computing, performance tuning, and automated testing and benchmarking. Earlier, across a year in data engineering and analytics, I built ETL pipelines and REST API integrations with Python and SQL, worked with Microsoft Azure, and delivered Power BI reporting for technical and non-technical stakeholders.

On the AI side, I build applied systems rather than experiments, including RepoFinder AI, which combines an LLM reasoning agent with a RAG-style retrieval pipeline, and real-time computer vision models in PyTorch.

HPC & Parallel Computing Applied AI / ML DevOps & Automation Data Engineering

Toolkit

Skills

Programming

  • Python
  • C/C++
  • Java
  • Scala
  • TLA+
  • LaTeX

AI / ML

  • Deep Learning
  • Computer Vision
  • NLP
  • Generative AI
  • PyTorch
  • TensorFlow
  • Spatio-Temporal Models
  • Model Deployment

HPC

  • Parallel Programming
  • GPU Computing
  • MPI, CUDA, OpenMP
  • Chapel
  • Job Scheduling
  • Accelerator Programming
  • SLURM
  • ReFrame & OSU Benchmarking
  • Profiling & Debugging on HPC
  • HPC CLI Environments

DevOps

  • Docker
  • Kubernetes
  • Terraform (IaC)
  • CI/CD
  • GitHub Actions
  • Git & Collaboration
  • Linux & Bash Scripting

Foundations

  • Networking & Distributed Systems
  • Computer Architecture
  • Software Engineering Principles
  • Parallel Algorithms & DS
  • Constraint Programming
  • QoS in TCP/IP
  • Routing & Flow Control

Data & Big Data

  • Hadoop & MapReduce
  • Apache Spark
  • Dataflow Automation & APIs
  • Data Engineering & Pipelines
  • Distributed File Systems
  • Geospatial Data Analysis
  • Data Cleaning & Management

Data Analytics

  • Excel
  • SQL
  • Power BI
  • Microsoft Azure
  • VBA
  • Python
  • ETL
  • Tableau
  • PostGIS / Geospatial

Web Development

  • HTML, CSS & JS
  • Node.js, Express & React
  • FastAPI, RESTful APIs
  • SQL & NoSQL
  • PHP
  • Authentication

Selected work

Projects

HPC regression testing benchmark results on Aion

HPC Regression Testing of MPI Communication

An automated regression testing suite to validate and monitor MPI communication performance on HPC clusters using ReFrame and OSU Micro-Benchmarks. Identified performance anomalies and optimized software stacks, ensuring a stable, high-performance communication fabric.

View on GitHub →
Real-time violence detection demo

Real-Time Violence Detection via Spatio-Temporal Deep Learning

Fine-tuned a state-of-the-art R(2+1)D-18 model to detect violent actions in video streams. Engineered data pipelines and training strategies to overcome overfitting and numerical instability, delivering real-time inference.

View on GitHub →
Real-time lip-reading demo

Real-Time Silent Speech Recognition

Designed a custom CNN+LSTM architecture in PyTorch to translate live lip movements into text in real-time. Implemented CTC loss, data preprocessing, and end-to-end deployment for a complete lip-reading system.

View on GitHub →
RepoFinder AI demo

RepoFinder AI: Intelligent GitHub Project Discovery

A full-stack search application that uses a dual-AI pipeline — a Llama-based reasoning agent and a Sentence Transformer — to translate natural language ideas into precise queries and discover relevant GitHub repositories.

View on GitHub →
Spacecraft detection and segmentation output

Spacecraft Recognition – Detection & Segmentation

Developed a two-stage computer vision pipeline for the SPARK 2024 challenge, combining YOLO11-Small detection with SegFormer-B1 segmentation to identify spacecraft body and solar panels in synthetic space imagery.

View on GitHub →
Housing Data Visualization output

Housing Data Visualization

Built an interactive data visualization dashboard using React and D3.js to explore housing market data across multiple dimensions. Implemented linked visualizations with real-time brushing and selection for intuitive pattern analysis.

View on GitHub →
CI/CD pipeline architecture

End-to-End CI/CD Pipeline for Full-Stack Application

Designed and implemented a complete CI/CD pipeline for a full-stack web application using GitLab, Docker, and Ansible. Automated build, testing, containerization, and deployment across development, staging, and production environments.

View on GitHub →
GitLab infrastructure automation diagram

GitLab CI/CD Infrastructure Automation

Built an end-to-end DevOps project that provisions a self-hosted GitLab server with Vagrant and Ansible and automates a 4-stage CI/CD pipeline for a Java Maven application using GitLab Runner and Docker.

View on GitHub →
VAST Challenge visual analytics platform

VAST Challenge 2022 – Visual Analytics Platform

Built a full-stack visual analytics application for VAST Challenge 2022 using a D3.js frontend, Node.js/Express backend, and PostgreSQL/PostGIS database to explore complex urban data through linked interactive views.

View on GitHub →
Confidential · available upon request

Extended Word Counting with Hadoop MapReduce

Built and benchmarked custom MapReduce jobs in Java for large-scale text processing, including regex-based filtering and performance evaluation on HPC clusters.

Private Repository
Confidential · available upon request

Word & Pair Counting with Spark RDDs

Reimplemented text analytics using Spark RDDs and Scala, leveraging in-memory computation to significantly improve performance over Hadoop-based approaches.

Private Repository
Confidential · available upon request

Titanic Dataset Analysis with Spark SQL

Analyzed passenger data using Spark DataFrames and SQL, implementing custom UDAFs and performing statistical analysis on structured datasets.

Private Repository
Confidential · available upon request

Marvel Social Network Graph Analysis

Constructed and analyzed a large-scale graph using Spark GraphX, applying PageRank and Connected Components to identify influential nodes and communities.

Private Repository
Confidential · available upon request

NYC Taxi Trips Geospatial Analysis

Processed large-scale geospatial data using Spark, performing borough-level analytics and anomaly detection on trip durations.

Private Repository
Confidential · available upon request

Financial Risk Analysis with Monte Carlo Simulation

Developed a distributed Monte Carlo simulation in Spark to compute financial risk metrics such as VaR and CVaR for stock portfolios.

Private Repository
Confidential · available upon request

Heart Disease Prediction (Classification)

Built a classification pipeline using Decision Trees and Random Forest with hyperparameter tuning via CrossValidator to predict heart disease risk.

Private Repository
Confidential · available upon request

Weather Prediction in Luxembourg (Regression)

Developed regression models using Spark MLlib to forecast temperature from time-series weather data with feature engineering and preprocessing.

Private Repository
Confidential · available upon request

Music Recommendation System (Collaborative Filtering)

Implemented an ALS-based recommender system with custom train/test splits and evaluated performance using AUC metrics.

Private Repository
Confidential · available upon request

Latent Semantic Analysis on Wikipedia

Performed topic modeling on 40K+ articles using LSA and cosine similarity to build a semantic document retrieval system.

Private Repository
Confidential · available upon request

Topic Modeling on Movie Plots

Applied NLP and LSA techniques on 134K movie plots, integrating metadata to enhance topic interpretation.

Private Repository
Confidential · available upon request

Outlier Detection on Sensor Data

Used K-Means clustering on 14M+ sensor records to detect anomalies and identify machine faults.

Private Repository
Confidential · available upon request

Big Graph Processing with OpenMP

Implemented a hybrid parallel BFS algorithm with dynamic strategy switching, achieving high performance on massive graphs.

Private Repository
Confidential · available upon request

Parallel Maze Pathfinding with MPI

Designed and benchmarked multiple MPI-based algorithms for distributed pathfinding, analyzing communication and scalability trade-offs.

Private Repository
Confidential · available upon request

Parallel Prefix-Sum with CUDA

Developed a GPU-based prefix-sum algorithm and used it to build efficient parallel data processing primitives.

Private Repository
Confidential · available upon request

Tiled Matrix Multiplication with CUDA

Optimized matrix multiplication using shared memory tiling in CUDA to significantly reduce memory latency and improve throughput.

Private Repository
Confidential · available upon request

Hand Raise Detection System (Real-Time CV)

Built a real-time detection system using OpenCV and MediaPipe to identify individuals raising hands, with custom heuristics and tracking logic.

Private Repository

Beyond the code

Professional Strengths

Stakeholder Communication

I communicate clearly with technical and non-technical stakeholders, helping translate ideas, requirements, and feedback into actionable work.

Technical Consultation

I enjoy understanding user needs, clarifying problems, and guiding discussions toward practical technical solutions.

Cross-Functional Collaboration

I work comfortably across research, operations, and engineering contexts, coordinating with different teams to keep work aligned and moving forward.

Presentation & Documentation

I can explain technical concepts in a structured way, present work clearly, and produce documentation that improves understanding and execution.

Ownership & Reliability

I take responsibility for assigned work, follow through carefully, and focus on delivering tasks with consistency, accuracy, and professionalism.

Problem Framing

I am good at listening, identifying the core issue behind a request, and helping connect the right people, tools, and next steps to solve it.

Career so far

Experience

  1. Mar 2026 – Aug 2026 · Full-time

    Software Development & Optimization Engineer

    PCOG, SnT – Interdisciplinary Centre for Security, Reliability & Trust · Supervised by Prof. Grégoire Danoy

    Design, develop, test and deliver performance-critical software in C++ and Python across the full development lifecycle — from requirements analysis through implementation, validation, automation and documentation.

    • Designed a production-quality optimisation solver in C++, translating complex requirements into structured, testable, maintainable architecture.
    • Delivered a 26.5× performance improvement on 28 cores through OpenMP multi-threading, profiling and bottleneck analysis; implemented GPU-accelerated components in CUDA.
    • Achieved 100% solution feasibility across all benchmark instances via systematic constraint handling and rigorous validation.
    • Executed a full architectural refactor verified as byte-identical to the original — zero functional regression.
    • Built automated testing, benchmarking and experiment pipelines for reproducible, repeatable results; analysed large experimental outputs into actionable findings.
    • Worked with Linux, Git, SLURM, containerized environments and scripted automation (Python, Bash); authored documentation and validation reports for technical and non-technical stakeholders.
  2. Mar 2025 – Feb 2026 · Part-time

    Junior Data Analyst

    SnT – Interdisciplinary Centre for Security, Reliability & Trust

    Supported data management, automation and reporting operations within the Technology Transfer Office — database administration, API integration, cloud-based workflows and process automation.

    • Maintained and validated relational databases and structured records, ensuring integrity and consistency across sources.
    • Automated data validation and processing workflows with Python, SQL and Excel/VBA; built REST API integrations and ETL dataflow pipelines between internal platforms.
    • Worked with Microsoft Azure and Power Platform for cloud-based workflows; built validation checks, monitoring and notification logic to catch data issues early.
    • Produced reporting, dashboards and KPI summaries with Power BI, Excel and Microsoft 365; acted as first point of contact for researchers on data submissions.
  3. Nov 2024 – Feb 2025 · Part-time

    Quality Assurance Test Engineer

    University of Luxembourg

    Tested, validated and documented Python-based command-line tooling used for research metrics evaluation.

    • Executed structured functional testing of Python CLI tools, verifying outputs against expected results and documenting defects.
    • Validated metrics outputs through systematic checks; provided actionable feedback improving usability, error handling and debugging output.
    • Improved technical documentation and strengthened reproducibility of analytical workflows.
  4. Sep 2023 – Aug 2024 · Full-time

    Software Developer

    Bizcope BD · Dhaka, Bangladesh

    Developed and maintained full-stack software solutions for web applications, mobile features and eCommerce/ERP systems, working directly from client and business requirements.

    • Built backend functionality with SQL-backed data operations; contributed to mobile features in cross-platform product delivery.
    • Collaborated with development, design and business stakeholders to design, build, test and deliver scalable digital products.
    • Supported the full SDLC from requirements and planning through implementation, deployment and maintenance, with documentation and code review.
  5. Sep 2021 – Jun 2023 · Student Employee

    Undergraduate Teaching Assistant

    North South University · Dhaka, Bangladesh

    Delivered tutorials and laboratory sessions for undergraduate programming courses in C and C++ across two academic years.

    • Taught C and C++ fundamentals in tutorials and hands-on labs for programming and data structures courses.
    • Guided students through debugging, algorithm design and problem-solving; assisted with course materials and assessments.
    • Evaluated assignments and mentored students with structured feedback on coding practice.

Academic path

Education

Sept 2024 – Present

M.Sc. High-Performance Computing

University of Luxembourg

  • Currently in the fourth semester of the HPC programme.
  • Doing the mandatory internship in the PCOG Research Group of SnT, Luxembourg.
  • Key learnings: Course details
Sept 2019 – Aug 2023

B.Sc. Computer Science & Engineering

North South University

  • Graduated with Magna Cum Laude distinction, ranked amongst the top of the class.
  • Active member of the NSU Problem Solvers club as a sports programmer, participating in competitive programming contests.
  • Key learnings: Course details

Let's talk

Get in Touch

If you'd like to collaborate, discuss an opportunity or just say hello, feel free to reach out!

jabinurmy07@gmail.com jabintasnimkhanurmy@gmail.com