Hi! I am a Senior Machine Learning Engineer at Spectrum Software & Consulting Ltd. (SSCL) and an AI researcher. Previously, I served as an Instructor, Research Assistant, and Undergraduate Teaching Assistant in the Department of Computer Science and Engineering at United International University (UIU), where I completed my B.Sc. in CSE, graduating 1st in my batch with Summa Cum Laude honors (CGPA: 3.97 / 4.00).

Feel free to reach out at phprogga@gmail.com for research collaborations or inquiries!

Research Interests
Multimodal Concept Learning Vision-Language Models (VLMs) Agentic AI & LLM Agents Computer Vision Natural Language Processing (NLP) Document AI & OCR Medical Image Analysis Explainable AI (XAI) Software Engineering

πŸ“ Publications & Manuscripts

Peer-Reviewed Publications
EACL 2026
Do Multi-Agents Solve Better Than Single?
Do Multi-Agents Solve Better Than Single? Evaluating Agentic Frameworks for Diagram-Grounded Geometry Problem Solving and Reasoning
Mahbub E. Sobhani*, Md. Faiyaz Abdullah Sayeedi*, Mohammad Nehad Alam, Proma Hossain Progga, and Swakkhar Shatabda
Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (EACL) — Student Research Workshop, 2026. ACL.
Neurocomputing '24
Siamese BiGRU Gait Recognition
A bidirectional Siamese recurrent neural network for accurate gait recognition using body landmarks
Proma Hossain Progga, Md. Jobayer Rahman, Swapnil Biswas, Md. Shakil Ahmed, Arif Reza Anwary, and Swakkhar Shatabda
Neurocomputing, Elsevier, 2024. (Q1 Journal)
IEEE I2CT '23
KNN Informative Instances
K-nearest neighbour classifier for big data mining based on informative instances
Proma Hossain Progga, Md. Jobayer Rahman, Swapnil Biswas, Md. Shakil Ahmed, and Dewan Md. Farid
International Conference for Convergence in Technology (IEEE I2CT), pp. 1–7, IEEE, 2023.
ICISS '23
UAVs in Green Health Care
UAVs in green health care for energy efficiency and real-time data transmission
Anika Khaer, Md. Siam Hossain Sarker, Proma Hossain Progga, Saniyat Mushrat Lamim, and Md. Motaharul Islam
International Conference on Intelligent Sustainable Systems (ICISS), pp. 773–788, Springer Nature Singapore, 2023.
ICONIP '22
iResSENet Retinal Blood Vessel Segmentation
iResSENet: An accurate convolutional neural network for retinal blood vessel segmentation
Proma Hossain Progga and Swakkhar Shatabda
International Conference on Neural Information Processing (ICONIP), pp. 567–578, Springer International Publishing, 2022.
Manuscripts Under Review
Under Review
DOCRELAY Multi-Agent Cross-Page Routing
DOCRELAY: Cross-Page Document Structure Extraction via Cost-Aware Multi-Agent Routing
F. H. Chowdhury, Proma Hossain Progga, M. F. A. Sayeedi, M. E. Sobhani, M. A. Arefeen, F. Y. Sadeque, M. F. Bari, and S. Shatabda
Under Review at NAACL 2027 (Also related to Skill-Routed Extraction and Assembly for Multi-Page Documents).

πŸ’Ό Professional & Academic Experience

Senior Machine Learning Engineer

Jul 2023 – Present
Spectrum Software & Consulting Ltd. (SSCL)

Leading the design, fine-tuning, and production deployment of enterprise Document AI platforms, multilingual Vision-Language Models (VLMs), Agentic AI workflows, digital onboarding (e-KYC) biometrics, trade compliance RAG engines, and real-time computer vision systems.

Instructor (Part-Time Lecturer)

Jun 2023 – Sep 2023
Department of CSE, United International University (UIU)

Delivered undergraduate lectures, designed course assessments, and mentored students in core computer science courses.

CSE 1115: Object-Oriented Programming (Summer '23) CSE 1325: Digital Logic Design (Summer '23)

Grader

Nov 2022 – May 2023
Department of CSE, United International University (UIU)

Evaluated assignments, programming projects, and coursework across AI, machine learning, and mathematics classes.

CSE 4889: Machine Learning (Spring '23) CSE 3811: Artificial Intelligence (Spring '23) CSE 1111: Structured Programming Language (Spring '23) MATH 2205: Probability & Statistics (Fall '22)

Undergraduate Teaching Assistant

Mar 2022 – May 2023
Department of CSE, United International University (UIU)

Assisted faculty in laboratory sessions, guided student software projects, and held regular consultation hours across four trimesters.

CSE 2216: Data Structures & Algorithms I Lab (Summer '22, Spring '23) CSE 3421: Software Engineering (Fall '22) CSE 3522: Database Management Systems Lab (Spring '22, Fall '22)

Research Assistant

Mar 2022 – Aug 2022
United International University (UIU)

Conducted research in deep learning, medical image segmentation, and pattern recognition under faculty supervision.

πŸš€ Selected AI & Research Projects

Research Agentic SE

GitAgent – Code-Aware Agentic Requirements Engineering

Researching an LLM-based agentic framework that transforms ambiguous stakeholder requests into implementation-ready engineering specifications through repository-grounded investigation and iterative human-in-the-loop requirements refinement.
Agentic AI LLM Agents Requirements Engineering Program Analysis Human-AI Interaction
Research Program Synthesis

Bug Fixing & Corner Test Case Generation

Explores automated bug detection and repair in code submitted during competitive programming, alongside generating edge and corner test cases to rigorously evaluate code robustness and functional correctness.
Program Synthesis LLM Competitive Programming Automated Testing
Research Multilingual NLP

Misspelled Banglish Text Correction

Addresses the challenges of normalizing and correcting code-mixed Banglish text (Bengali and English) that frequently exhibits phonetic variations and inconsistent transliterated spellings.
NLP LLM Multilingual AI Code-Mixed Text
Research Graph + VLM

Table Structure Recognition & Semantic Mapping

Focuses on recognizing complex tabular structures within financial trade-related documentsβ€”identifying cell boundaries, multi-line headers, and structural relationships between rows and columns.
GNN Object Detection OCR LLM Multimodal
Industry Lead Enterprise Platform

XtractAI – Enterprise Document AI Platform

  • Led development of an enterprise Document AI platform for classification, OCR, and structured information extraction.
  • Integrated LLMs and Vision-Language Models into scalable inference pipelines with production APIs for real-time PDF and scan processing.
Document AI Vision-Language Models LLM OCR Information Extraction
Industry 97.48% EN / 91.23% BN

Identra-KYC – AI-Powered Digital Onboarding (e-KYC)

  • Designed an AI-powered onboarding platform integrating OCR, VLMs, face matching, and video liveness detection.
  • Fine-tuned multilingual VLMs achieving 97.48% (English) and 91.23% (Bangla) NID extraction accuracy.
  • Generated 10K+ synthetic NID samples and integrated Bangladesh Election Commission APIs for real-time verification.
Vision-Language Models OCR Face Biometrics Liveness Detection Multimodal
Industry 82% Cost Reduction

Import Document Scrutiny & Trade Compliance

  • Automated validation of LC MT700 clauses against import documents using GPT-OSS-20B, DotOCR, and RAG.
  • Implemented compliance verification based on UCP 600 and ISBP 821, reducing manual effort and operational cost by 82%.
LLM RAG DotOCR Trade Compliance Document AI
Industry 3–5 s/page Latency

SpectraTrade – Financial Document Intelligence

  • Fine-tuned and developed VLMs and layout-aware OCR for 24+ financial document types with 3–5 s/page latency.
  • Deployed production AI services for PrimeBanijjo, Bangladesh's first AI-powered digital LC platform.
Vision-Language Models LLM OCR Document Classification Multimodal
Industry 86% Cost Reduction

DockStream – RMG Automation Platform

  • Built end-to-end AI and RPA pipelines for document classification, information extraction, and automated web form submission.
  • Reduced manual effort and operational cost by 86% in high-volume RMG workflows.
OCR Information Extraction RPA NLP Document AI
Industry 200 ms CPU Inference

Bangla License Plate Detection & Recognition (ALPR)

  • Developed a production-grade ALPR system deployed across multiple toll plazas in Bangladesh.
  • Designed a transformer-based OCR architecture replacing CNN approaches, achieving end-to-end detection and text extraction within 200 ms CPU-only inference.
Computer Vision Transformers OCR Object Detection Real-Time Inference
Industry Layout-Aware AI

Identity & Financial Document Parsing

  • Developed AI pipelines for extracting and validating information from Bangladeshi NIDs, passports, bank cheques, EXP forms, and fixed-layout financial documents.
  • Built OCR- and layout-aware extraction systems for structured and semi-structured documents.
Layout Analysis OCR Information Extraction Computer Vision
Industry Autonomous Agents

Agentic AI for Finance Automation & SQA Testing

  • Developed Agentic AI workflows for automated QA test-case generation, execution, and software validation.
  • Built LLM agents for automating finance-related workflows, improving operational efficiency and reducing manual intervention.
Agentic AI LLM Agents Workflow Automation Software Testing
Industry Smart Surveillance

Anomaly Detection & Smart Parking Analytics

  • Developed computer vision-based anomaly detection and parking analytics solutions for automated monitoring and intelligent decision-making.
  • Built real-world AI pipelines for multi-object detection, tracking, and abnormal event identification.
Computer Vision Object Detection Multi-Object Tracking Anomaly Detection
University Research Bioinformatics

Graph Neural Network to Analyze Protein Secondary Structure

Utilized Graph Neural Networks (GNNs) on the ccPDB 2.0 dataset to predict protein secondary structure and benchmarked performance against SVM, AdaBoost, KNN, and Decision Tree models.
GNN TensorFlow SVM / AdaBoost / KNN Bioinformatics
Kaggle Competition 5th Place Winner

Bengali Automatic Speech Recognition (ASR)

Built a state-of-the-art speech recognition pipeline for out-of-distribution Bengali audio using a 400-hour crowdsourced dataset (Common Voice), integrating PyTorch acoustic models with BERT, Word2Vec, and 6-gram language models.
PyTorch BERT Word2Vec N-Gram Speech Recognition
URC Mars Rover 1st in Asia

Biological Life Detection through Rock Classification

Developed the planetary science subsystem to categorize soil and rock samples into extant, extinct, or 'No Presence of Life' using custom datasets and CNN/VGG deep learning architectures.
CNN VGG Deep Learning Mars Rover (URC)

πŸ› οΈ Technical Skills

Programming Languages

Python C++ C Java PHP SQL

AI & ML Frameworks

PyTorch TensorFlow Keras LangChain LangGraph OpenCV spaCy

Backend & APIs

FastAPI Flask Django REST APIs Microservices

Database & Retrieval

PostgreSQL MySQL MongoDB Vector Embeddings Knowledge Retrieval (RAG)

MLOps & Deployment

Docker Git GitHub GitLab CI/CD Linux Model Deployment ML Pipelines

Web Development

HTML5 CSS3 / SASS JavaScript Bootstrap

πŸŽ“ Education

B.Sc. in Computer Science and Engineering

Jul 2019 – May 2023
School of Engineering, United International University (UIU), Dhaka, Bangladesh

CGPA: 3.97 out of 4.00 (Summa Cum Laude)
Batch Position: 1st (Department of CSE)
Undergraduate Thesis: Human Identification using Gait Landmarks

πŸ† Honors, Certifications & Leadership

πŸ… Summa Cum Laude & 1st in Batch (Dept. of CSE)
United International University
Graduated at the top of the class with a CGPA of 3.97 / 4.00.
2024
πŸŽ“ 100% Tuition Fee Waiver Scholarship
United International University
Awarded 100% merit scholarship in 10 out of 12 trimesters, reserved for the top 2% of students.
2019 – 2023
πŸš€ 1st Place in Asia & 13th Globally β€” University Rover Challenge (URC)
Mars Society (International Robotics Competition)
Designed and built the Mars rover science & life-detection subsystem for simulated Martian exploration.
2022
πŸ₯ˆ 1st Runner-Up β€” CSE Project Showcase
United International University
Led the development of Protibaad, a public safety and community awareness platform.
2022
πŸŽ–οΈ 5th Position β€” Bengali Automatic Speech Recognition (Kaggle Competition)
DL Sprint – BUET CSE Fest
Recognized among top teams nationwide for out-of-distribution Bengali speech recognition on 400 hours of audio.
2022

Certifications & Workshops

  • Machine Learning Specialization (2023)
    Stanford Online & DeepLearning.AI (Coursera)
  • Technical Workshop Instructor (2022 – 2023)
    Conducted multiple hands-on workshops on Java, Image Processing, and related topics at United International University (UIU).

Extracurricular Leadership

  • Senior Executive, Research & Development (2021 – 2023)
    UIU Robotics Club
  • Programming Wing Executive (2021 – 2023)
    UIU Computer Club
  • Instructor (2020 – 2023)
    UIU App Forum