Quazi

Work

Technical Specialist — Computer Vision & AI

Transoft Solutions Inc | Canada (Remote)
Jan 2025 - Present

I lead applied R&D at the intersection of computer vision and agentic AI, defining system architecture, model strategy, and the long-term AI roadmap for enterprise transportation-safety products. My current work spans multi-agent LLM systems, retrieval-augmented generation, and large-scale perception pipelines built for production reliability.

Key Responsibilities:

  • Architect and ship agentic AI systems — multi-agent orchestration, tool-using LLM workflows, and autonomous pipelines for automated engineering-data analysis, technical-document synthesis, and report generation, reducing manual reporting effort by 70%.
  • Design production-grade RAG (Retrieval-Augmented Generation) pipelines with embedding optimization, vector databases, hybrid retrieval, re-ranking, and evaluation harnesses for high-accuracy, context-aware, low-hallucination responses.
  • Drive LLM fine-tuning and model training workflows — domain adaptation, instruction tuning, prompt and context engineering, and rigorous offline/online evaluation before deployment.
  • Build transportation-focused computer vision systems — object detection, semantic segmentation, classification, and multi-object tracking — achieving 90%+ mAP in production under real-world constraints.
  • Own the full MLOps lifecycle across multiple AI products: problem framing, data strategy, labeling standards, experimentation, benchmarking, deployment automation, observability, and post-deployment monitoring.

Senior Backend / AI Engineer

Deal Analysis Software Intelligence (DASI) — via Toptal | Remote
Jun 2025 - Nov 2025

Led the architecture and deployment of a serverless AI commerce platform that transforms user room images into structured furniture intelligence and delivers real-time product recommendations across Amazon, eBay, Walmart, and Google Shopping.

Key Responsibilities:

  • Directed integration of LLMs and multimodal models (ChatGPT, Hugging Face) for furniture description generation, combined with marketplace scraping APIs optimized for reliability, latency, and cost.
  • Designed and shipped an advanced image inpainting feature for interactive furniture selection and replacement, leveraging computer vision and cloud-based generative AI.
  • Built secure, subscription-ready payment infrastructure with Stripe, including checkout flows and backend monetization logic.
  • Refactored monolithic scraping services into modular, scalable microservices, improving maintainability and cross-marketplace support.

Computer Vision & AI Team Lead

Advanced Mobility Analytics Group | Brisbane, Australia
Feb 2021 - Dec 2024

As Team Lead of the Computer Vision and AI team, I managed a high-performing team and drove innovation in applied deep learning and computer vision for proactive road safety.

Key Responsibilities:

  • Led architecture and delivery of three production AI video-analytics platforms, contributing to the prevention of 692+ crashes and 113+ injuries across deployed regions.
  • Architected scalable AWS video-processing infrastructure (Batch, SageMaker, Lambda, Step Functions, S3, EC2) capable of analyzing 1,000+ hours of video per day, reducing turnaround time by 40%.
  • Designed and deployed real-time multi-object detection and tracking systems on edge devices supporting 20+ road-user classes, improving violation and conflict detection accuracy by 15%.
  • Owned the full ML lifecycle — data strategy, experimentation, deployment, and monitoring — across detection, segmentation, pose estimation, and tracking systems in production.
  • Led cross-functional teams across CV, engineering, QA, and delivery, driving platform-level performance improvements of 20%.

Senior Machine Learning Engineer

TigerIT Bangladesh Ltd | Dhaka, Bangladesh
Apr 2014 - Feb 2018

As a Computer Vision and Deep Learning Engineer, I designed and shipped advanced perception systems, pushing the boundaries of real-time recognition.

Key Responsibilities:

  • Designed an end-to-end Bengali license-plate detection system using computer vision and deep learning, enabling real-time speed estimation and traffic monitoring.
  • Developed an ICAO-compliant face verification system with Scikit-Learn and Python, and designed a faster facial-recognition algorithm using 10% less memory than state-of-the-art approaches.
  • Led R&D of a palm-print detection algorithm using deep neural networks with autoencoders, outperforming the previously deployed system by 10% while running 5% faster.
  • Collaborated with cross-functional research and engineering teams to validate approaches and deliver production systems.

Software Engineer

KAZ Software Limited | Dhaka, Bangladesh
Oct 2013 - Apr 2014

Full-stack engineer on a web-based knowledge-graph aggregation platform that collected and organized information from the web based on user queries.

Key Responsibilities:

  • Built front-end and backend for the platform using KnockoutJS and NodeJS, storing and querying a Neo4j knowledge graph.
  • Designed scalable RESTful APIs and optimized graph queries for efficient relationship traversal and real-time search.
  • Managed the full software lifecycle — planning, development, testing, deployment, and maintenance — in a client-facing environment.

Software Engineer

Nascenia IT Limited | Dhaka, Bangladesh
Jun 2012 - Sep 2013

Designed and deployed web-based products using Ruby on Rails, Django, and JavaScript, leading development from architecture to production.

Key Responsibilities:

  • Led a team building full-stack web products from conceptualization to delivery.
  • Designed RESTful APIs for internet-connected home-appliance (IoT) controller systems, enabling secure device communication and remote management.
  • Participated in project planning, system architecture, and database design; maintained development, staging, and production environments for high reliability and data integrity.
  • Conducted code reviews ensuring maintainability and scalability of web applications.