DePaul Research Labs

Graduate Research Assistant | Jan 2024 – Present

• Conducted research in multimodal fusion, vision-language models, and transformer interpretability.
• Published 2 papers and presented findings at internal symposiums and ML reading groups.
• Developed recommendation systems for personalized content delivery, improving user engagement by 20%.
• Built NLP models for text classification and sentiment analysis, achieving 95% accuracy on benchmark datasets.
• Designed and implemented computer vision models for image recognition, reducing inference latency by 25%.
• Conducted data analysis on large-scale datasets using Hadoop and Spark, enabling data-driven decision-making.

IBM

AI/ML Engineer | Febuary 2021 – August 2023

• Led development of production-grade generative AI pipelines and agentic frameworks (CAF, AutoGPT).
• Deployed post-training optimization techniques (RLHF, quantization) to reduce model latency by 40%.
• Built scalable services using Rust and AWS Inferentia for multimodal and LLM-based applications.
• Built and optimized recommendation systems using collaborative filtering and deep learning models, processing 1TB+ of user interaction data daily.
• Developed NLP models for text generation and summarization, improving model accuracy by 10%.
• Automated model deployment pipelines using Docker and Jenkins, reducing deployment time by 40%.
• Conducted performance tuning for ML models, leveraging PyTorch to improve efficiency and scalability.
• Collaborated with cross-functional teams to deliver AI solutions aligned with business goals.

Technocolabs

AI/ML Intern | November 2020 – December 2020

• Implemented k-anonymity pipelines for sensitive order data (Python/Spark), reducing PII leakage by 83%.
• Optimized BigQuery SQL for search log analysis, reducing feature generation time from 4hrs to 90 minutes.
• Built multi-model ensemble (XGBoost + Isolation Forest) detecting
• 92% of defective shipments (vs. 78% manual inspection baseline).
• 2.4x faster than legacy systems through TensorRT optimizations.
• Developed latency prediction service (Django + TensorFlow Lite).
• Achieved 37% reduction in inference variability critical for real-time ETA systems.
• Processed 50+ RPM with less than 300ms latency (simulating peak demand scenarios)

Machine Learning Fundamentals

Supervised & Unsupervised Learning

Regression, Classification, Clustering, Dimensionality Reduction

Reinforcement Learning

Q-Learning, PPO, DDPG

Ensemble & Bayesian Methods

Random Forest, XGBoost, Naive Bayes

Deep Learning

Architectures

CNNs, RNNs, Transformers

Optimization & Regularization

SGD, Adam, Dropout, BatchNorm

Generative Models

GANs, VAEs, Diffusion Models

Natural Language Processing

Language Models

BERT, GPT, T5, LLaMA

NLP Tasks

NER, Sentiment Analysis, Machine Translation

Vector Databases

FAISS, Pinecone, Milvus

Computer Vision

Detection & Segmentation

YOLO, Faster R-CNN, U-Net

Video & Multimodal

3D CNNs, Optical Flow, CLIP, Flamingo

MLOps & Production

Deployment & Cloud

Docker, FastAPI, AWS, GCP, Azure

Monitoring & CI/CD

Prometheus, Jenkins, GitHub Actions

Tools, Math & Specialized Areas

Libraries & Big Data

PyTorch, TensorFlow, Spark, Hadoop

Math & Statistics

Linear Algebra, Probability, A/B Testing

Specialized Domains

Time Series, GNNs, Recommenders, Edge AI