// Hello, I'm
Production Computer Vision · Generative AI (RAG & AI Agents) · GPU-Optimized MLOps
I design, build, and own production AI at national scale — computer-vision systems serving 2M+ retail outlets, plus current-generation RAG, multi-agent, and real-time voice systems, all on GPU-optimized, self-hosted infrastructure.
I'm an Applied AI Engineer with 3+ years designing, building, and deploying production AI at national scale. At HawkEyes Digital Monitoring I own applied-AI delivery end to end — from research and model design through deployment and monitoring — across retail-execution, manufacturing, and surveillance product lines, and I mentor a growing AI team.
My core is deep computer vision (YOLO detection/segmentation, OpenCV) shipping at ~99% accuracy across 2M+ outlets for enterprise FMCG and mobile-financial-services clients — BAT, Unilever, and Nagad among them. I pair that with current generative-AI delivery: production RAG with vector search and LangGraph, self-correcting multi-agent systems, and real-time voice agents — all backed by GPU-optimized async inference and containerized MLOps.
I build and run my own infrastructure: two on-premises Ubuntu GPU servers I provisioned and administer host most of the team's AI services and training jobs, cutting reliance on rented cloud GPUs. I'm currently building a custom Bangladeshi-accent Bangla text-to-speech model (VITS + grapheme frontend) to power production AI voice agents for inbound/outbound call automation. I care equally about model accuracy and clean, observable production engineering.
Computer Vision
YOLO detection & segmentation at 2M+ outlets, ~99% accuracy.
Generative AI
Production RAG, LangGraph multi-agent systems, voice agents.
MLOps & Infra
Self-hosted GPU servers, Docker, GPU-optimized inference.
Dhaka, Bangladesh
HawkEyes Digital Monitoring Ltd.
B.Sc. in Electrical & Electronic Engineering
Bangladesh University of Business and Technology (BUBT)
2021 · CGPA 3.75 / 4.00
Bangla (native) · English (professional working proficiency)
Grouped by what I actually ship — each cluster is backed by real, in-production projects, not self-rated percentages.
Detection, segmentation & analytics shipped at retail scale
Grounded, self-correcting LLM systems in production
Low-resource Bangla speech & document intelligence
Async-first services engineered for throughput
In-house GPU infra and reproducible deployment
Foundations and the Linux environment it all runs on
Eight deep case studies — the strongest of 55 repositories across 8 AI domains. Each follows problem → approach → architecture → results.
Dhaka, Bangladesh
Dhaka, Bangladesh
Remote
Real-Time Custom Object Detection (Upwork)Interested in production AI, computer vision, or LLM/agent systems? I'm open to BD and international / remote roles and collaboration.
