Where the work has been.
The tracks I have actually shipped in, in the order I grew into them — classical machine learning first, then deep learning and vision, then generative AI, and now whole-system architecture around all three.
- Building AI
- 4.9+years, since 2022
- Industry roles
- 4across Bangladesh and the US
- Currently at
- BanglalinkAI & Automation Engineer
Role by role
Each position with what it actually involved. Open any role to read the full detail.
- Jan 2026 – Present11 mosCurrent role
AI & Automation Engineer (Full-Time, Augmented)
Banglalink Digital Communications Ltd., Gulshan 1, Dhaka
Architecting Blink, Banglalink's enterprise AI assistant, with advanced retrieval and full production observability.
- Hybrid search with Reciprocal Rank Fusion
- 100% of AI interactions traced
- Bilingual Bangla/English, multimodal
- Architecting and developing Blink, an advanced RAG-based AI assistant for Banglalink with customised retrieval pipelines, hybrid search, multimodal support, auto skill selection and bilingual (Bangla/English) capability.
- Implemented end-to-end LLM observability and tracing with Langfuse, tracking 100% of AI interactions across all production pipelines and cutting debugging time.
Read more (2)Show less
- Built a real-time performance dashboard in Grafana on PostgreSQL for operational insight into the assistant's usage and accuracy.
- Retrieval combines keyword and dense vector search with Reciprocal Rank Fusion (RRF) for more accurate, grounded answers.
- OpenAI
- LangChain
- LangGraph
- Qdrant
- FastAPI
- Langfuse
- Grafana
- PostgreSQL
- Docker
- Aug 2025 – Nov 20254 mos
AI Software Engineer (Remote)
Corsult, Valley Stream, NY 11581, United States
Automation workflows and a real-time AI web call assistant for a US team, delivered remotely.
- Real-time AI web call assistant
- Social media automation with GPT-4
- Developed social media automation workflows using n8n/Make.com integrated with OpenAI's GPT-4 API.
- Built an AI-Based real-time Web Call Assistant (OpenAI, WebRTC) who can conversation with users on web calls.
Read more (2)Show less
- Built an employee ID card system with Vibe Coding and prompt-engineering
- Worked remotely from Bangladesh, collaborating with US teams and automating operations.
- OpenAI GPT-4
- n8n
- Make.com
- WebRTC
- Jan 2025 – Aug 20258 mos
AI Engineer (Contract)
Daffodil International University, Dhaka
Built and deployed AI products for the university, from a bilingual voice assistant to face-search photo finding.
- AI Proctor: 24/7 chat + voice
- Champion, DIU National Project Showcasing
- Designed and deployed AI Proctor, a bilingual (Bangla/English) chat + voice assistant for the university's Proctor Office.
- Built an AI-Based Photo Finder using face recognition and vector similarity search to help users locate their images in large photo albums.
Read more (2)Show less
- Led the development of DIU Smart Transport, a digital campus mobility system.
- Ongoing Projects: Currently engaged in R&D on Data Analysis Agent, AI-Powered Photo Finder, and Virtual Try-On Assistant.
- OpenAI
- LangChain
- FAISS
- FastAPI
- LiveKit Agents
- face_recognition
- June 2024 – Jan 20258 mos
AI Developer (Project Basis)
Center for Data Science and Research (CDSR), Bangladesh | Remote
Research-driven AI for education and vision: handwritten OCR for UNESCO, RAG chatbots and plate detection.
- OCR for UNESCO's Rohingya learners
- RAG chatbot on open-source LLMs
- Developed a handwritten OCR system using Microsoft's TROCR model for UNESCO's online education platform targeting Rohingya children.
- Integrated a RAG-based chatbot using Mistral LLM and ChromaDB for document-based question answering.
Read more (2)Show less
- Built a YOLOv8-based vehicle number plate detection system with OCR support.
- Worked on multiple LLM-based applications using open-source models (LLaMA, Mistral via Hugging Face).
- TrOCR
- Mistral
- ChromaDB
- YOLOv8
- Hugging Face
- February 2022 – Present4 yrs 10 mos
AI/ML Journey Started
Self-Learning & Exploration
Where it started: self-taught machine learning that grew into production AI engineering.
- Author of “Trailblazing into Machine Learning”
- DoctorHimel_V1 fine-tuned LLM on Hugging Face
- Began my journey in machine learning driven by a passion to explore how data and algorithms can solve real-world problems.
- Over the past four-plus years, I've grown into an AI Engineer with hands-on experience in Deep Learning, Computer vision, GenAI, LLMs, Agentic AI and backend development.
Read more (2)Show less
- Work spans from building experimental prototypes to leading end-to-end development of production-ready AI systems.
- Always focused on delivering scalable, impactful solutions through intelligent automation and thoughtful design.
- Python
- scikit-learn
- TensorFlow
- PyTorch
- OpenCV
Practice areas
Four tracks that overlap rather than replace each other. I still take classical machine learning work when it is the honest answer to the problem.
2022 – Present
Machine Learning Model Development
Designed and deployed ML models for real-world prediction tasks like heart disease detection, customer churn, and salary prediction. Expertise in scikit-learn, hyperparameter tuning (GridSearchCV, K-Fold), and statistical data analysis.
2022 – Present
Deep Learning & Computer Vision Projects
Built and trained CNN & ANN models using TensorFlow and PyTorch. Delivered OCR systems, disease detection models, and real-time object tracking with YOLO, image segmentation, transfer learning, and SMOTE augmentation.
2023 – Present
Generative AI & LLM Engineering
Built advanced AI apps using OpenAI, Hugging Face, DeepSeek, and Langchain. Specialized in RAG systems, vector search (FAISS, ChromaDB), agent frameworks (CrewAI, LangGraph), and fine-tuning LLMs with LoRA/QLoRA.
2023 – Present
AI System Architecture & Solution Design
Designed scalable end-to-end AI systems with multi-agent collaboration, API orchestration, and ethical AI practices. Mapped business use-cases into modular AI solutions using modern stack.
Have an AI problem worth solving?
Tell me what you are trying to build. I will tell you honestly whether AI is the right tool, what it would take, and how I would approach it.
Typical reply within 24 hours
Working from Dhaka, Bangladesh · remote worldwide