About Me
Hello, my name is Ali. I have extensive experience designing, deploying, and operating production AI/ML systems at scale — from multi-agent LLM architectures to conversational AI handling millions of calls a day. I have authored several research publications and hold a dual master's degree.
I am an AI/ML Engineer by day and a Mad Scientist by night (evil laugh intended), always chasing the most state-of-the-art algorithms. I care about coding, mathematical modeling, optimization, LLMs, multi-agent systems, machine learning, and the intersection of all these.
Work Experience
AI/ML Engineer
Capgemini Client: Charter Spectrum — Kansas City, KS
May 2026 – Present
- Keep LLM-driven conversational AI agents running in production on Kore.ai — supporting roughly 1M IVR calls a day for Charter Spectrum’s customer service operations. When that many people are on the line, latency and NLU quality are not optional.
- Leading the technical migration off Kore.ai onto an in-house conversational AI platform — architecting the NLU, dialog management, and orchestration layers that replace the third-party stack. The payoff: eliminating platform/licensing costs for a projected $4M in annual savings.
- Partnering with Google on an NGA initiative to bring natural, human-like generative voice to IVR — working the tradeoffs between model selection, latency, and voice quality so the bot sounds like a person, not a hold-music menu.
Founding AI/ML Engineer
Apply Sprint Acquired — Remote
Aug 2025 – May 2026
- Went from zero to production building a multi-agent AI system (LangGraph + LangChain) that matches candidates to jobs and rewrites their resumes — all orchestrated by GPT-4, Claude, and DeepSeek working in concert. Under the hood: a two-stage NLP pipeline with a fine-tuned XLM-RoBERTa classifier (PyTorch) feeding a stacked ensemble, the whole thing living in Dockerized containers on AWS. Wanna see? Check it out here.
- Wore a lot of hats as Founding Engineer — designed the architecture, shaped the ML pipeline, shipped to prod, and kept a small team of engineers pointed in the right direction, all the way through acquisition.
Senior Data Scientist
Bank of America — Charlotte, NC
Feb 2024 – Aug 2025
- Kept a watchful eye on fraud models guarding $20B+ in annual transactions across Mastercard and OMNI credit card portfolios — tracking CSI, PSI, AUC-ROC, and rank ordering to make sure nothing drifted off the rails (regulators included).
- Dug deep into dozens of KPIs to find the "why" behind model misbehavior, and built surrogate fraud detection models using black-box distillation to reverse-engineer what third-party models were actually up to.
- Translated gnarly model performance findings into clean executive reports that senior leadership could actually act on — no jargon overload, just clear insights and a path forward.
Data Scientist
Lowe's — Charlotte, NC
Feb 2022 – Jan 2024
- Media Mix Modeling: Built a statistical framework to figure out which marketing dollars were actually doing anything. Used XGBoost as a baseline, wrestled with multicollinearity, and layered in Bayesian and ensemble methods for clean attribution. The payoff? A 3% ROI uplift on a $900M marketing budget — real money.
- Predicting Sales During Storms: When hurricanes hit, Lowe's needs to know exactly what to stock and where. Built regression models to forecast sales across 3,000 stores and 30,000 product assortments during severe weather — using K-Means clustering to find patterns, and a full Docker + Kubernetes + MLflow pipeline to ship it all cleanly to prod.
Data Scientist
HTN Networks Inc. (Cisco) — Irvine, CA
Apr 2020 – Jan 2022
- Built an ML system to hunt down malicious network traffic — combining reinforcement learning with an ensemble model that kept getting smarter through continuous experimentation. Set up data drift monitoring so it wouldn't go stale after deployment.
- The numbers speak for themselves: 97% F1 score, 42% fewer false alarms, and 35% faster threat response. The network security team was happy. The attackers, less so.
Graduate Research Assistant
NC State University — Raleigh, NC
Aug 2017 – Jan 2018
- Joined a sustainability research lab and put ANOVA and ARIMA to work — forecasting environmental trends and squeezing more reliability out of the predictions used in impact assessments.
- Turned messy environmental datasets into actionable insights using good old-fashioned statistics. Unglamorous, important work.
Teaching Assistant
Istanbul Sehir University — Istanbul, Turkey
Sep 2015 – May 2017
- Helped undergrads survive their coursework — grading, tutoring, and running discussion sessions without making anyone cry (most of the time).
- Ran Python and SQL labs that actually made students better coders. Turns out explaining the same bug twelve different ways is a great way to really understand it yourself.
Education
Master's in Operations Research
North Carolina State University
Raleigh, NC
Master's in Industrial & Systems Engineering
Istanbul Sehir University
Istanbul, Turkey
B.E. Industrial & Manufacturing Engineering
NED University
Karachi, Pakistan
Honors & Awards
NC State PhD Fellowship — Recognized with the NC State PhD fellowship award, a prestigious award granted to a select few pursuing a PhD, recognizing exceptional academic excellence and research potential.
NC, USA
Things I Build for Fun
Wisdom Bot (NLP) — answers questions with quotes from wise people. Try it here!
GAN Art — Neurostyling experiments. See the painting below! On a totally unrelated note, I’m having a cup of tea with my sister in Sedona, Arizona — nice change of scenery while I tinker with this stuff.
Music Generation — Currently learning to generate music via GAN. Coming soon!
Outdoors — Soccer, hiking & exploring beautiful places.
Selected Projects
Bitcoin Price Forecasting
Multi-variate forecast of Bitcoin prices using LSTM neural networks.
View ProjectGrocery Item Classifier
Classifier based on pretrained VGG-16 + CNN layers to distinguish 25 grocery products. Achieved 90% accuracy on unseen data.
View ProjectCycle GAN Data Augmentation
Implemented Cycle GANs to double image dataset size. Generated data improved model performance by 10%.
View ProjectAircraft Maintenance Scheduling
Optimized aircraft maintenance schedules using Mixed Integer Linear Programming (MILP) via CPLEX.
View ProjectCapacitated Vehicle Routing (VRP)
VRP challenge by vorto.ai — an autonomous supply chain platform.
View ProjectNetwork Intrusion Detection
ML models to detect malicious network activity that compromises security.
View ProjectNASA Battery Life Prediction
Predicting remaining battery life of NASA's small electric aircraft using Neural Networks.
View ProjectTime Series: ARIMA & LSTM
Time series analysis comparing ARIMA and LSTM models on healthcare data.
View ProjectNLP: Quora Question Classification
Identify and flag insincere questions using LSTM, CNN, and GRU architectures.
View Project