I build AI products end-to-end — from LLM workflows and applied ML models to backend APIs, data pipelines, integrations, and observability.

Most of my recent work has been around production LLM systems: agentic workflows, RAG, AI-powered editing/generation pipelines, evaluation, and multi-tenant SaaS infrastructure. I also have a research background across healthcare AI, EEG/time-series modeling, NLP, and computer vision.

I like working on AI systems that are technically solid, product-aware, and reliable enough to be used outside a notebook.

• Production LLM applications and agentic workflows
• LangGraph/LangChain orchestration, tool use, structured outputs, and human-in-the-loop review
• RAG systems, prompt/tool design, evaluation, and workflow reliability
• Backend services and APIs for AI products
• Multi-tenant SaaS workflows, authentication, integrations, and data pipelines
• Applied ML research in healthcare AI, EEG, NLP, computer vision, and time-series modeling

• Built agentic AI workflows for content generation, editing, creative variation, draft validation, and human-in-the-loop review.
• Worked on LangGraph-based conversational flows with state management, checkpointing, and multi-turn product workflows.
• Built AI generation and publishing pipelines across image generation, AI editing, scoring, and campaign publishing.
• Developed backend APIs and orchestration layers for AI workflows, third-party integrations, and frontend-ready product experiences.
• Shipped platform features across authentication, OAuth/RBAC, workspace management, billing, and multi-tenant SaaS workflows.
• Improved production reliability through validation, observability, testing, and workflow-level debugging.

• Built Transformer/LSTM pipelines for EEG-based cerebral edema prediction.
• Worked on healthcare AI, clinical ML, biomedical knowledge graphs, and retrieval workflows.
• Ran deep learning experiments across time-series, NLP, and clinical datasets using PyTorch/CUDA and GPU clusters.

• ISS Earth Imagery Geolocation — geolocation system for astronaut photography using computer vision and multimodal techniques.
• Conversational Stock Investment Advisor — NLP assistant for stock-related queries using Rasa, NER, BERT, sentiment analysis, and financial APIs.
• Neural Style Transfer Study — comparative study of neural style transfer models.
• Real-Time Drone Detection — object detection system using CNN, YOLO, and SSD models.
• Biased Prosecution Project — applied ML/NLP project focused on legal and social-impact analysis.

I’ve published and contributed to work across healthcare AI, EEG modeling, computer vision, NLP, and applied machine learning.

Selected publications:

• Automated Cerebral Edema Detection using EEG in Post-Cardiac Arrest Patients
• Geolocating Earth Imagery from ISS: Machine Learning for Astronaut Photography Mapping
• A Comparative Study of Neural Style Transfer Models
• Drone Detection using YOLO and SSD: A Comparative Study