AI & Data Pipeline Engineer · NYC
Data Engineer for Financial Systems
I'm Abir, an AI & Data Pipeline Engineer focused on credit risk modeling, explainable AI, and cloud-native ETL architecture.
Previously: Fixnox, HKexcel Education. MS Financial Technology from UConn.
Experience
Professional experience
AutoShopIQ
AI & Data Engineer
Sept 2025 – Present · Stamford, CT
- Architected an event-driven ingestion pipeline on AWS (S3 event triggers, Lambda, IAM, Textract) that captures, parses, and stores repair orders for analytics and ML.
- Built a Python capture agent that uploads invoices to S3 under a configured prefix, files them by RO number, and prints paper copies, with retries and failure quarantine so printing survives AWS outages.
- Diagnosed a Windows Session 0 printing failure and re-architected the deployment as a hidden Scheduled Task, verified on a live shop PC through reboot.
- Cut print time from ~243 s to ~2.4 s by benchmarking and bundling a PDFium-based print helper, eliminating the Acrobat dependency.
- Shipped a no-Python, seven-file installer package for shop PCs, published to GitHub with deployment runbooks.
- Secured the pipeline with least-privilege, write-only IAM credentials and PII-safe S3 object keys that exclude customer names.
- Deployed an Amazon Textract Lambda parser via the AWS CLI through test, scorecard, production, and canary stages, validated on 10 large invoices.
- Root-causing parser defects across ~9,000 repair orders using Textract geometry analysis and a replay harness built on saved raw JSON.
Fixnox
Data Engineer Intern
Jan 2024 – Dec 2024 · Sydney, Australia (Remote)
- Architected a batch ETL pipeline on AWS (S3, Glue, Athena) to ingest and analyze FIX protocol trade messages following a medallion architecture.
- Developed a Python-based FIX message data generator producing ~23,000 realistic trade logs partitioned in Hive-style format.
- Engineered ETL transformations with PySpark and Pandas to parse raw CSV trade logs into optimized Parquet with derived trading metrics.
HKexcel Education
Physics & Mathematics Instructor
Nov 2019 – Jul 2023 · Hong Kong
- Delivered instruction to 70+ IB Mathematics and Physics students, resulting in consistent scores above 80%.
- Translated complex concepts into digestible modules, increasing student comprehension and achievement by 25%.
City University of Hong Kong Solar Car Team
Mechanical Engineer
Jun 2017 – Sep 2020 · Hong Kong SAR
- Served as a core mechanical engineer focused on CAD design and body fabrication for the CityUHK Solar Car Team.
- Contributed to building a semi-electric solar car completed in September 2018 and participated in the Sasol Solar Challenge 2018.
Aarhus University
Summer Research Intern
Jun 2018 – Aug 2018 · Denmark
- Worked on improving mechanical design, calibration, and controller design for a robotic exoskeleton prototype targeting Adolescent Idiopathic Scoliosis (AIS).
- Supported six degrees-of-freedom Stewart platform control using approaches ranging from PD control to Adaptive Inverse Dynamic and impedance control.
Department of Electronic Engineering, City University of Hong Kong
Student Intern
Jul 2017 – Mar 2018 · Hong Kong SAR
- Contributed to polymer waveguide device design, fabrication, and characterization for integrated optical chips used in communication, sensing, and bio/chemical applications.
Projects
Featured work
Credit Risk Scoring Engine with Explainable AI
End-to-end credit risk classifier using XGBoost on 10,000 synthetic credit records with SHAP explainability for regulatory transparency.
- ROC AUC of 0.77 on 2,000-record test set
- SHAP waterfall plots, force plots, and beeswarm visualizations
- Addresses ECOA, FCRA, GDPR Article 22 requirements
- Interactive Streamlit dashboard with real-time scoring
Monte Carlo Portfolio Optimization
Simulated buy-and-hold vs rebalancing strategies (2017–2022), improving risk-return tradeoff through monthly rebalancing with S&P 500 data.
- Analyzed stock investment strategies and risk-return tradeoffs
- Monthly rebalancing improved portfolio performance
Predicting Problematic Internet Use
Classification model for adolescent internet risk using the HBN dataset from the Child Mind Institute.
- Built on the Healthy Brain Network (HBN) dataset
- F1 score of 0.73 for internet addiction risk prediction
Education
Academic background
University of Connecticut
Aug 2023 – May 2025 · Storrs, CT
MS Financial Technology
Key Coursework
City University of Hong Kong
Sept 2015 – Oct 2019 · Hong Kong
BEng Mechatronic Engineering
Certifications
Associate Data Engineer in SQL
DataCamp
Python & Statistics for Financial Analysis
Coursera
Python for Data Science, AI & Development
Coursera
Multivariate Calculus for Machine Learning
Coursera
Skills
Technical stack
Tools and platforms I use most in data engineering and ML work.