Saim Mehmood

• Engineered autonomous AI agents to analyze advisor-representative consultations, extracting compliance insights to improve client service quality and regulatory adherence

• Automated contact enrichment pipeline using Playwright and Node.js to scrape and verify financial advisor data from regulatory databases and Google, reducing manual contact research time by 50% for the wealth management team

• Containerized AI agent services using Docker and Amazon ECR, deploying on EC2 instances to ensure reliable, cost-effective availability for on-demand workloads

• Architected serverless workflows via AWS Lambda to execute complex AI agent prompts, complementing container-based deployments for lightweight, on-demand tasks

• Contributed to rubric-driven AI model training, focusing on reasoning quality, instruction adherence, and robustness

• Built structured ML Jupyter notebooks from prompt → starter code → final reference implementation

• Designed datasets and evaluation criteria to systematically test model performance and failure modes

• Designed Data Pipelines to extract, transform and load (ETL) data from customer information systems (CIS) into meter data management (MDM) software using Rules Engine (proprietary ETL tool) and SQL

• Managed the processing, transformation and extraction of meter data coming from advanced metering infrastructure (AMI) systems such as (Sensus, Itron, L+G) using Multispeak Web Services (SOAP/WSDL, REST APIs)

• Implemented modules such for line loss analysis, leak detection and transformer load analysis, reducing utility losses by (10-20%) and improving operational efficiency - data solutions & data models

• Collaborated with internal stakeholders and clients to define business requirements and deliver tailored solutions - compliance, service delivery

• Provided technical support and troubleshooting throughout project implementation - project management, configuration management

• Contributed to the development and integration of a proprietary LLM-based analytics module, extending MDM insights - Python, PowerBI, LLM, google-t5 (text-to-text transformer based model)

• Built end-to-end machine learning models (training, deployment & evaluation) using AutoML and custom model training by leveraging Vertex AI, BigQuery, Docker and Dataflow for a client in financial space. Improving their ability to train ML models more efficiently using GCP - Python, ML Libraries (scikit-learn, Tensorflow/Keras/PyTorch), pandas, numpy

• Developed and deployed semantic models to detect bias in appraisal documents, leveraging Document AI and Cohere API

• Certified Google Cloud Professional Machine Learning Engineer URL: https://bit.ly/3WhLBzg

Taught multiple courses at the Electrical Engineering and Computer Science (EECS) department

• Courses: Programming for Mobile Computing EECS 1022 - Introduction to Database Systems EECS 3412 - Object Oriented Programming from Sensors to Actuators - EECS 1021 - Software Design - EECS 3311

Tasks include: directing tutorials, exam invigilation, final and midterm exam review sessions, grading assignments/exams, office hour duties. OOP, Java, Android Studio, IntelliJ

• The research was related to trajectory data mining, machine learning, and statistical inference

• Developed a method that utilizes trajectories of [cars, pedestrians, etc.] as a way to infer semantic similarities between geographical areas

• Published the research titled Learning Semantic Relationships of Geographical Areas based on Trajectories at the IEEE Mobile Data Management Conference 2020 Versailles, France, and received Best Paper Award

• Designed a framework inside SNIC using Apache Spark, SparkR & Jupyter Notebook to simplify computations of highly parallel scientific applications

• Our project titled Towards Moving Scientific Applications in the Cloud enabled researchers to seamlessly deploy their applications on the spark server and scale it to multiple worker nodes as needed