Saleh Ahmed Shafin

AI Engineer specializing in agentic AI systems – architecting multi-agent orchestration platforms, LLM/RAG pipelines, and automation tools with persistent memory, business-document intelligence, and multi-format content generation.

• Location: Nikunjo-2, Dhaka, Bangladesh
• Email: salehshafin.ahmed@gmail.com
• Phone: +880 1832115427
• Portfolio: SalehAhmedShafin.github.io
• LinkedIn: linkedin.com/in/saleh-ahmed-shafin-0773b0193
• Kaggle: kaggle.com/positivesas
• Codeforces: codeforces.com/profile/Sa_Shafin
• LeetCode: leetcode.com/sa_shafin

💼 Experience

AI Engineer - Codemen Solutions Limited
December 2023 - Present

• Architected Alex, a multi-agent Lead Orchestrator that deconstructs single or multi-step, multi-file natural-language requests and autonomously routes them across 16 specialized skill agents, serving 52 SaaS and enterprise clients alongside a 4-member global engineering team.
• Built business-document intelligence skills for Alex: automated Tenant COI compliance gap analysis between lease agreements and insurance certificates, invoice/utility-bill data extraction, lease abstraction, and equipment-manual-based preventative-maintenance scheduling, cutting manual document review time by 80-85%.
• Built Project Intelligence and Prototype Intelligence pipelines that convert natural-language requirements into BRD/SAD documents and then directly into downloadable, functional software prototype ZIP files, plus citation-backed Document Q&A (RAG over indexed files), cutting requirement-to-prototype turnaround by 85-90%.
• Built content-generation and research skills for Alex: Media-to-Documentation and File Intelligence (structured document and custom data-point extraction from files, video, and audio), multi-format Document Generator (DOCX/XLSX/CSV/JSON/PDF/HTML), AI image generation, automated slide-deck creation, and Research Intelligence for real-time web data.
• Developed Lore, a long- and mid-term memory knowledge base giving agents persistent context retention across 1200+ sessions, improving multi-turn task accuracy/relevance by 80-90%.
• Built an Automated Purchase Order (PO) Processing agent that extracts, validates, and processes purchase orders end-to-end, reducing manual processing time by 90-95% and processing 36000+ POs to date.
• Built the Energy Advantage System, an agent-driven automation that ingests multiple input files and consolidates them into a multi-sheet Excel output, replacing 5-6 hours of manual reporting per cycle with a 4-5-minute automated run.

Digital Marketer - EGEN
Jan 2022 - Feb 2023

• Gained skills in SEO, content marketing, social media management, PPC advertising, data analysis, and more for effective online product promotion.

ITEE Training Program Level 2 - BD ITEC
March 2023 - April 2023

• Learned the fundamentals of information technology and gained skills for success in the IT industry.

🛠️ Skills

• Core AI: Agentic AI, Multi-Agent Systems, LLM, LangChain, LlamaIndex, RAG (Vector/Graph/Hybrid), FlowiseAI, ELSA Workflow, NLP, Computer Vision, XAI.
• Frameworks: PyTorch, TensorFlow, Keras, Scikit-learn, OpenCV, Django, FastAPI.
• Programming Languages: Python, C/C++, Java, JavaScript, PHP.
• Tools & Cloud: GitHub, Docker, AWS.
• Databases: VectorDB, Neo4j, ArangoDB, MySQL, Firebase, MSSQL, Oracle, PLSQL.
• Soft Skills: Problem Solving, Communication, Self-learning, Adaptability, Reliable and Consistent.

🎓 Education

• Bachelor of Science in Computer Science and Engineering

  • Ahsanullah University of Science and Technology

    2019-2023

    CGPA: 3.488

• Higher Secondary Certificate (HSC)

  • Dhaka City College

    2015-2017

    GPA: 5.00

• Secondary School Certificate (SSC)

  • Tarash Islamia Pilot High School

    2010-2015

    GPA: 5.00

🔬 Publications

Explainable Lung Disease Classification from Chest X-Ray Images Utilizing Deep Learning and XAI

• Publication Link
• Description: This research focuses on accurately classifying various lung diseases from chest X-ray images using advanced deep learning architectures. To ensure clinical reliability and transparency, Explainable AI (XAI) techniques were integrated, allowing medical professionals to interpret and trust the model’s decision-making process.
• Technologies/Focus: Deep Learning, Explainable AI (XAI), Medical Image Processing, Computer Vision.

🏆 Contests and Achievements

• Nascenia AI Hackathon - Bengali Medical Dialogue Generation: Position: 4th
• ML Olympiad - CO2 Emissions Prediction Challenge: Position: 4th
• Robi Datathon 3.0: Position: 23rd
• BUET DL Sprint 2.0: Position: 16th
• DU DataVerse Challenge - ITVerse 2023: Position: 10th
• IUT Bengali.AI: ASR for Regional Dialects: Position: 5th
• Intra AUST Programming Contest: Spring’20 (Position: 15th), Codecombat1.0 (Position: 9th)
• Bengali.AI Speech Recognition: Position: 360th
• Zilla Science Fair (Project - Generation of electricity without fuel): Prize: 1st (April 2013)
• Scholarships: Khwaja Yonus Ali Scholarship, Talent Pool Scholarship

💻 Academic Projects

1. Ensemble GPT: Text and voice chat with LLM using Streamlit UI and LangChain

• GitHub Link
• Description: Enables querying document-specific and external information through text and voice using a Streamlit UI and LangChain integration.
• Technology Used: LLM, LangChain, Streamlit, API (Google, OpenAI), voice-to-text, text-to-voice.

2. Multimodal Disaster Event Identification from Social Media Posts

• GitHub Link
• Description: Identifies disaster-damaged materials from social media posts by merging text and image data.
• Technology Used: Python, PyTorch, NLP, Computer Vision.

3. XAI and Deep Neural Networks for Crop Disease Detection and Interpretability

• GitHub Link
• Description: Integrates Explainable AI techniques with Deep Neural Networks for crop disease detection.
• Technology Used: Python, TensorFlow, Keras, Computer Vision, XAI.

4. Segmentation of a large multi-domain Bengali document layout analysis dataset

• GitHub Link
• Description: This project means breaking down a big collection of different types of Bengali documents into smaller parts based on their layout. This process helps organize and categorize the documents, making it easier to work with them for various purposes, like searching, analyzing, or extracting information.
• Technology Used: Vision Transformer, Segmentation.

5. Depth Estimation Of Endoscopy Using Pytorch

• GitHub Link
• Description: It refers to the process of using the PyTorch deep learning framework to determine the depth or 3D information in endoscopy images or videos. This can be valuable in medical applications, as it helps in understanding the spatial characteristics of structures inside the body when using endoscopic cameras.
• Technology Used: PyTorch, Deep Learning.

6. Bangladesh Real Estate Market Analysis

• GitHub Link
• Description: Exploration of a comprehensive house listing dataset from various cities in Bangladesh for real estate market insights.
• Technology Used: Python, Pandas, Scikit-learn, XGBoost, LightGBM, Random Forest.

7. Bangladesh Economic Growth Analysis

• GitHub Link
• Description: Explore and visualize data showcasing Bangladesh’s robust economic growth, highlighting its position as one of the world’s fastest-growing economies.
• Technology Used: NumPy, Pandas, Matplotlib, Seaborn, Scikit-learn, pmdarima, folium.

8. Broadband Management System

• GitHub Link
• Description: Streamlined broadband service management tool built using Java Swing and MSSQL.
• Technology Used: Java, JavaSwing, MSSQL.

9. Shop Management with Distribution Database System

• GitHub Link
• Description: Insert products into different site databases and introduce searching functionality from different sites.
• Technology Used: C++, PLSQL, Oracle.

10. Bachelor-Neer (House rental management system)

• GitHub Link
• Description: Online platform for house rental management, providing various accommodation options.
• Technology Used: MVC, .Net Framework, JavaScript, HTML, CSS.

11. Jersey Lagbe

• GitHub Link
• Description: “Jersey Lagbe” in Bengali means “I need a jersey.” This is Jersey selling website which is developed by HTML, CSS, Bootstrap, Carousel, JS, PHP.
• Technology Used: HTML, CSS, Bootstrap, Carousel, JS, PHP.

12. Easy Tour

• GitHub Link
• Description: This is android Studio based project. In this software anyone can see the tourist place in bangladesh also take a small guideline from this application.
• Technology Used: Android Studio, Java, Firebase Database.

13. Automated Water Pump with pH

• GitHub Link
• Description: The Arduino turns the water pump ON/OFF by calculating the data received from the Ultrasonic Distance Sensor. Then measure PH of this water using PH-meter then abnormal value aware using buzzer.
• Technology Used: Arduino, Hardware.

14. Ludo game

• GitHub Link
• Description: This project explores the development and implementation of Ludo games using WebGL (Web Graphics Library) technology, enabling the creation of immersive and visually appealing digital versions in a web browser environment.
• Technology Used: WebGL, Three.js.

15. PyGame- Snake

• GitHub Link
• Description: It is Basic Snake Game in Python! Navigate around the screen, trying to eat food and grow in length. The game utilizes the Pygame library for graphics and game logic.
• Technology Used: Python, VSCode, Pygame.

16. Captain White Apron Fighting Game

• GitHub Link
• Description: 2D desktop game featuring a doctor combating the COVID-19 pandemic by eliminating viruses.
• Technology Used: C, C++, iGraphics.