LLM Product Attribute Generation
Fine-tuned and deployed a Mistral 7B language model pipeline on AWS to generate product attributes, improving accuracy by 30%.
Prior role · E-commerce data · 2024
Highlights from our founder's prior roles and personal projects. Case studies from Northbeam AI client engagements will be added here as they are completed.
Results from production systems built in prior employment. Employers are described rather than named.
Fine-tuned and deployed a Mistral 7B language model pipeline on AWS to generate product attributes, improving accuracy by 30%.
Prior role · E-commerce data · 2024Built and deployed a real-time fraud detection system with Apache Kafka and Spark on Snowflake data, processing millions of transactions.
Prior role · Transaction data · 2023Automated monthly revenue calculations and built interactive Tableau dashboards, cutting calculation time for management by 93%.
Prior role · Business reporting · 2024Worked with engineering teams to forecast demand and power production for a solar farm, improving prediction accuracy by 40%.
Prior role · Energy sector · 2018 to 2021Open-source projects with full code on GitHub, several with a written walkthrough.
Streams transactions through Kafka and flags suspicious activity with Spark as it happens.
A walkthrough of Kafka pipeline architecture and implementation, from producers to consumers.
Extracts weather data, transforms it and loads it into PostgreSQL on a scheduled Airflow workflow.
Uses Apache Spark to clean, transform and analyze large volumes of customer data.
A complete extract, transform and load process built entirely in SQL.
Trains and compares several machine learning models to find the most effective fraud detector.
Classifies book genres by analyzing the sentiment of reader reviews and feedback.
Compares classification models for predicting diabetes from patient health indicators.
For new projects, book a free 30-minute call or email shahrukhshl0@gmail.com.
Book a CallTampa, Florida
Working remotely with clients nationwide