Arsalan Q.

I build AI applications that turn documents and support data into useful answers, classifications, and working software. My focus is GenAI, NLP, and Python APIs, from data preparation and model evaluation to application integration and testing. WHAT I CAN HELP YOU BUILD • RAG and document Q&A: retrieval pipelines, source-grounded answers, and evaluation against your own documents. • AI assistants and agents: structured extraction, tool calling, MCP-connected tools, and integrations with applications such as Slack. • NLP and machine learning: text classification, preprocessing, baseline models, and task-specific evaluation. • ML-backed applications: FastAPI inference endpoints, batch processing, Streamlit dashboards, and Docker packaging. RELEVANT PROJECT EXPERIENCE TriageIQ — Built a support-ticket classification system with a FastAPI service, batch inference, a Streamlit interface, and MLflow tracking. The TF-IDF/logistic-regression intent model recorded 97.9% accuracy on a held-out split of the 27-intent Bitext dataset. This is a dataset-specific evaluation result. Verity — Built a Slack fact-checking agent using Gemini and Brave Search through MCP. It extracts claims from text and links, retrieves supporting sources, and returns structured, cited verdicts. OpportunityPulse AI — Built a GenAI application for structured opportunity extraction and application drafting, with deterministic profile matching, a React interface, and Supabase integration. SafeX Solutions — Served as an AI/ML and Automation Intern and group leader. My contributions included a retrieval-based FAQ assistant and a support-chatbot prototype combining live LLM APIs with a local TF-IDF fallback. I also work on GlycoTwin, an academic capstone involving forecasting experiments, model comparison, a FastAPI backend, and a retrieval-based bilingual AI coach. TOOLS Python, scikit-learn, PyTorch, Hugging Face, LangChain, Gemini/OpenAI APIs, FastAPI, Streamlit, Docker, and MLflow. I can help with a focused prototype, an existing AI feature that needs improvement, or an ML model that needs an application interface. I work from a clear scope, evaluate the output, and provide setup notes and a practical handover. Send me the problem, a representative input, and the output you need.