Senior AI/ML Engineer GenAI & App Modernization

Amir Jamali

+1-438-***-**** ******.****@*****.***

LinkedIn: amir-jamali GitHub: amir-jamali

Montreal, Quebec, Canada

Senior AI & Machine Learning Engineer Computer Vision Generative AI LLMs PROFESSIONAL SUMMARY

Senior AI & Machine Learning Engineer with 10+ years of experience building production-scale AI systems across computer vision, deep learning, and multimodal AI. Proven expertise in developing real-world perception systems, including detection, segmentation, and large-scale data-driven model training. Experienced in designing and deploying LLM-powered applications and prompt engineering workflows, including structured prompting, few-shot design, prompt optimization, and automated evaluation to improve quality, consistency, latency, and cost. Built retrieval-augmented generation (RAG) and agent-based workflows and delivered scalable, API-first AI services using Python and FastAPI. Strong track record of owning end-to-end AI solutions from problem definition and modeling to production deployment and monitoring.

EXPERIENCE

• Rival Solutions (rivalsolutions.com) Apr 2022 – Present Senior AI & Machine Learning Engineer Montreal, Canada

Owned the architecture and delivery of production-grade ML and computer vision systems for large-scale infrastructure and road-condition analytics.

Designed end-to-end pipelines spanning data ingestion, labeling workflows, training, evaluation, deployment, and continuous monitoring.

Built API-first inference services supporting computer vision, multimodal models, and GenAI workflows; ensured strict schemas, versioned artifacts, and reproducible deployments.

Established MLOps workflows including experiment tracking, model registry practices, CI/CD automation, and evaluation/regression testing.

Led technical decision-making around model architecture, performance optimization, reliability, and cost-efficiency in real-world production settings.

Collaborated cross-functionally with engineering, product, and domain stakeholders to translate requirements into scalable AI systems.

Provisioned and managed cloud infrastructure using Infrastructure as Code (Terraform), including Kubernetes clusters, storage, and networking resources.

Developed production-grade AI and GenAI applications in Python, building reusable libraries and scalable service architectures.

Designed and maintained production prompt engineering workflows for large language models (OpenAI, GPT-4, Claude), including structured prompting, few-shot templates, and domain-specific instruction design.

Implemented structured outputs (JSON schema / function-style responses) with validation and fallback logic to ensure reliable downstream processing.

Built automated prompt evaluation pipelines using benchmark datasets to measure quality, consistency, latency, and cost across prompt and model versions.

Performed prompt and model comparisons (A/B-style) and introduced prompt versioning and regression testing to maintain performance across model updates.

Designed and deployed Retrieval-Augmented Generation (RAG) pipelines including document ingestion, chunking, embedding, retrieval, reranking, and citation-aware response generation.

Designed and productionized scalable PySpark-based data pipelines for finance and insurance, enabling behavioral risk prediction, anomaly detection, and fraud analytics using distributed feature engineering, aggregation strategies, window functions, and temporal modeling on large-scale datasets.

• L’ACADEMIE (lacademiepro.ca) Jan 2026 – Present

AI & Machine Learning Instructor(Part-Time) Montreal, Canada

Teaching practical AI, Computer Vision, and Machine Learning concepts through hands-on projects.

Mentoring students in Deep Learning, Generative AI, LLMs, and Retrieval-Augmented Generation (RAG) systems.

Guiding implementation of AI applications using Python, PyTorch, OpenCV, and LangChain-based workflows.

• Faraadid Vision Feb 2014 – July 2019

Senior Computer Vision & Machine Learning Engineer Tehran, Iran

Designed and deployed real-time computer vision systems for surveillance, safety, and monitoring applications operating on live camera feeds.

Built end-to-end vision pipelines for detection, tracking, recognition, and event analysis under constrained and noisy real-world conditions.

Led development of fire and smoke detection systems deployed in large-scale outdoor environments with kilometer-range visibility.

Implemented face recognition, people counting, intrusion detection, and license plate recognition systems used in production settings.

Optimized vision pipelines for performance and reliability using OpenCV, C++, GPU acceleration, and system-level tuning.

Acted as technical lead for multiple projects, mentoring junior engineers and coordinating delivery across hardware, software, and deployment teams.

• K. N. Toosi University of Technology Feb 2008 – Nov 2012 Electronics Engineer / Hardware Designer Tehran, Iran

Designed schematics and PCB layouts for high-power 1kW FM transmitter power supply systems.

Developed analog, DSP, and power electronic circuits for RF and communication hardware applications.

Worked on signal processing and signal-to-noise ratio (SNR) optimization for transmitter and communication systems.

Performed circuit simulation, PCB routing, hardware debugging, and system validation.

Collaborated on integration and testing of transmitter hardware components and power regulation systems. SELECTED TECHNICAL EXPERIENCE & SYSTEMS

Computer Vision & Multimodal Perception

• Designed and deployed deep learning pipelines for image classification, semantic segmentation, and instance segmentation using CNN- and Transformer-based architectures.

• Built real-world perception systems for road and urban environments, including detection and localization of pavement defects, lane markings, and traffic signs.

• Applied geometric computer vision techniques including camera calibration, projection models, vanishing point estimation, and monocular 3D reasoning for spatial understanding.

• Developed tooling for spherical-to-cubic image transformations and polygon annotation mapping across projection domains.

• Generated synthetic image datasets for computer vision training, including traffic signs, pavement defects, occlusions, and shadow artifacts to improve model generalization and reduce false positives.

• Fine-tuned diffusion-based generative models to produce prompt-controlled and mask-guided synthetic images for data augmentation and robustness testing.

• Integrated OCR, super-resolution, and post-processing pipelines to improve recognition accuracy on small, noisy, or degraded visual inputs.

• Built a CrewAI-based multi-agent vision system where a vision–OCR agent (GPT-4o-mini) extracts text from traffic sign images and a domain-specific MUTCD agent (GPT-4.1) interprets sign type, category, and official MUTCD codes, producing structured, consistently formatted metadata.

• Developed high-performance computer vision pipelines by building OpenCV with CUDA and offloading compute-intensive operations to GPU-accelerated kernels. ML & GenAI Systems

• Designed and implemented end-to-end machine learning and GenAI systems from data ingestion through model training, evaluation, deployment, and monitoring.

• Customized deep learning architectures, including multi-task learning pipelines with shared representations and uncertainty-aware loss fusion.

• Fine-tuned diffusion and generative models for image synthesis, prompt-conditioned generation, and mask-guided augmentation.

• Built multimodal learning systems combining vision, text, audio, and structured data.

• Designed and optimized LLM workflows including structured prompting, automated evaluation, and performance optimization.

Retrieval, Agents & Reasoning

• Designed and implemented Retrieval-Augmented Generation (RAG) pipelines, including document ingestion, chunking, embedding, indexing, retrieval, reranking, and citation-aware generation.

• Built conversational, agentic, and multimodal RAG systems supporting complex reasoning workflows.

• Designed multi-agent AI systems using LangGraph- and CrewAI-style orchestration patterns.

• Implemented planner, executor, critic, and verification agents with bounded retries, deterministic execution graphs, and safety controls.

• Integrated vision- and OCR-based agents with reasoning pipelines for semantic interpretation and validation. Data Engineering & Big Data

• Built large-scale batch data ingestion and feature engineering pipelines using PySpark and Spark SQL.

• Designed data-lake-style architectures for raw, processed, and materialized datasets.

• Implemented distributed joins, aggregations, window functions, data validation, and ML-ready dataset generation.

• Generated synthetic tabular, text, and image datasets with distribution validation and lineage tracking. Audio, Speech & Multimodal Systems

• Developed automatic speech recognition (ASR) pipelines for speech-to-text transcription.

• Implemented speaker embedding, enrollment, and verification systems.

• Built text-to-speech (TTS) pipelines with safety controls, rate limiting, and monitoring.

• Integrated audio pipelines into multimodal systems combining vision, language, and metadata. Model Serving, MLOps & State Management

• Designed API-first model inference services using FastAPI with strict schema validation.

• Implemented unified serving layers for ML, GenAI, and agent-based systems.

• Integrated experiment tracking, artifact versioning, and model registry workflows using MLflow-style systems.

• Built CI/CD pipelines and automated testing for model training and deployment.

• Implemented structured logging, metrics collection, drift detection, alerting, and evaluation pipelines.

• Integrated Redis for caching, state management, agent coordination, rate limiting, and execution control. Deployment, Infrastructure & Performance

• Containerized ML and AI services using Docker and Docker Compose.

• Designed and deployed Kubernetes-based systems with autoscaling, rollout strategies, and health checks.

• Implemented canary deployments, shadow routing, and A/B testing for production inference systems.

• Optimized performance-critical pipelines using CUDA-enabled OpenCV and GPU-accelerated components.

• Designed scalable execution patterns with retries, cancellation, rollback, and failure isolation. EDUCATION

• Concordia University Sep 2020 - May 2023

M.Sc. in Computer Science Montreal, Canada

GPA: 3.77/4.00

Thesis: Pavement Defect Classification and Localization Using Hybrid Weakly Supervised and Supervised Deep Learning and GIS

Proposed a two-stage framework combining Class Activation Mapping (CAM) and supervised segmentation

(U-Net/Mask R-CNN) for efficient road defect detection, trained with GIS-based datasets and achieving up to 97% precision on crack types.

• Amir Kabir University of Technology (Tehran Polytechnic) Sep 2012 - Apr 2015 M.Sc. in Computer Engineering Tehran, Iran

Grade: 17.07/20

Thesis: Sparse and Dense 3D Face Modelling based on 2.5D AAM Approach from a Single Image

Proposed a modified 2.5D AAM-based framework to estimate 3D face pose and reconstruct dense 3D face models from a single image, using camera calibration for pose recovery and enhancing recognition with 3DLBP features.

• K. N. Toosi University of Technology Sep 2001 - May 2006 B.Sc. in Electronics Tehran, Iran

Grade: 16.03/20

Thesis: Three Phase Three Level Matrix and Diode Clamp DC/AC Inverter PATENTS AND PUBLICATIONS C=CONFERENCE, J=JOURNAL, P=PATENT, S=IN SUBMISSION, T=THESIS

[C1] Jamali, A., Laflamme, C., Hammad, A., & Huber, R. (2023). Pavement Defect Classification and Localization Using Weakly Supervised Deep Learning. In Proceedings of the Creative Construction Conference (CCC), Budapest, Hungary, July 2023.

SKILLS

• Programming Languages: Python, C, C++17, MATLAB, Bash

• Core Machine Learning: scikit-learn, XGBoost, LightGBM, Feature Engineering, Model Evaluation, Statistical Validation

• Deep Learning Frameworks: PyTorch, TensorFlow, Keras

• Computer Vision:OpenCV, CUDA, GPU-Accelerated OpenCV, Detectron2, YOLOv8, Faster R-CNN, Mask R-CNN, U-Net, DeepLabV3+, SegFormer, SAM, SAMURAI Tracker

• Vision-Language & Multimodal Models: CLIP, BLIP, SigLIP, Vision Transformers (ViT), Multimodal Embedding Models

• Large Language Models (LLMs) & GenAI: LLaMA, Mistral, Qwen, Hugging Face Transformers, LangChain, LangGraph, CrewAI, Prompt Engineering, Prompt Optimization, Prompt Evaluation, Few-Shot Prompting

• LLM Runtimes & API Providers:Ollama, OpenAI API, Anthropic Claude, Google Gemini

• Diffusion & Generative Models: Stable Diffusion, ControlNet, DreamBooth, LoRA, Text-to-Image, Image-to-Image, Inpainting

• Audio & Speech Systems:Whisper (ASR), Speaker Embeddings, Speaker Verification, Text-to-Speech (TTS), Audio Feature Extraction

• Retrieval-Augmented Generation (RAG): FAISS, Dense Embeddings, Reranking, Conversational RAG, Agentic RAG, Multimodal RAG

• Big Data & Distributed Processing:Databricks, PySpark, Spark SQL, Hadoop, DataFrame API, Distributed Joins, Window Functions, Batch Feature Engineering, Data Lake Architectures

• MLOps & Model Lifecycle:MLflow, Model Registry, Experiment Tracking, Dataset Versioning, Drift Detection, A/B Testing, Canary & Shadow Deployments

• API Development & Model Serving: FastAPI, Flask, RESTful API Design, Pydantic, Model Serving Architectures

• State Management & Caching: Redis (Caching, Rate Limiting, Pipeline State, Agent Memory)

• DevOps & CI/CD:Docker, Docker Hub, Docker Compose, Kubernetes, Terraform (IaC), GitHub Actions, GitLab CI, Jenkins

• Cloud Platforms:AWS, Google Cloud Platform (GCP), Microsoft Azure

• Databases & Storage: PostgreSQL, MySQL, MongoDB, SQL-Based Feature Tables, Object Storage Concepts

(S3-style), Data Lake Layouts

• Annotation & Dataset Tooling: CVAT, Labelme, Roboflow, VGG Image Annotator, Custom OpenCV-Based Annotation Tools

• Frontend & Visualization:Angular, React, REST-Based UI Integration, Monitoring Dashboards

• Embedded Systems & Hardware:NVIDIA Jetson, Raspberry Pi, UART, SPI, I2C, Power Electronics

• Developer Tools:Git, GitHub, GitLab, Linux, Django, VS Code, PyCharm, Jupyter Notebook, Shell Scripting, LaTeX

HONORS AND AWARDS

• Mitacs Accelerate Research Grant (2021) — Concordia University in collaboration with Rival Solutions. Successfully secured a Mitacs Accelerate project fostering collaboration between academia and industry to develop advanced AI-driven pavement evaluation systems.

• Professional Project Recognition (2016) — Department of Environment, Iran. Fire & Smoke Detection Developed and successfully tested a real-time fire and smoke detection application at the Department of Environment, capable of detecting smoke from IP camera feeds at distances of up to 2 km. Implemented optimized computer vision algorithms for rapid detection and integrated an automated SMS alert system using a GSM module via a serial port to notify relevant authorities.

• Top 0.5% in National M.Sc. Entrance Exam (2012) — Ministry of Science, Research and Technology, Iran. Ranked among 30,000 candidates based on academic excellence and analytical skills.

• Top 0.3% in National B.Sc. Entrance Exam (2001) — Ministry of Science, Research and Technology, Iran. Ranked among 260,000 candidates in a highly competitive national assessment. LEADERSHIP EXPERIENCE

• Team Lead – Computer Vision Group (2017 – 2019) — Faraadid Vision Co., Iran Led a team focused on developing computer vision–based surveillance applications for real-time monitoring and threat detection. Oversaw the design and implementation of intelligent systems for fire and smoke detection, intrusion detection, people counting, and license plate recognition. Coordinated project planning, team task allocation, and system integration across multiple platforms. CERTIFICATIONS

• Deep Learning Specialization (deeplearning.ai 2018)

• Course: Neural Networks and Deep Learning

• Course: Improving Deep Neural Networks

• Course: Structuring Machine Learning Projects

• Course: Convolutional Neural Networks

• Course: Sequence Models

LANGUAGE

• English: Fluent

• Persian:Native

Contact this candidate