Hi, I'm Jens
Creative and curious mind, passionate about leading teams, solving hard problems, and driving impact.
Aspiring to roles that sit between strategy and execution and translating vision and business goals into models, metrics, and real workflows. I combine technical literacy with structured problem solving to decide when to trust a model, when to challenge it, and how to embed tools into decisions people actually make.
Projects
MicroPilot — On-Device Sign Detection
Fine-tuned a 0.1B multimodal LLM (MiniMind-O) with LoRA to detect US speed limit and stop signs from Tesla dashcam footage at 91.9% accuracy. Deployed on a Raspberry Pi edge device in a 2014 Toyota Corolla for real-time offline inference; proof of concept for a larger Advanced Driver Assist System (ADAS).
UW Foster MBA — AI Course Pilot
A RAG-powered course planning assistant for UW Foster MBA students. Recommends electives based on career goals by searching across 49 real course syllabi, and predicts when each course is historically offered using past quarterly schedules.
Multi-Agent Chatbot System
A complex conversational AI project featuring persistent chat history and community moderation capabilities built in Python. This multi-agent architecture was designed to securely and effectively handle varied user intents while self-moderating.
PythonAI Stock Analysis Agent
A chat-based AI agent powered by Claude that autonomously merges technical stock analysis (training a custom PyTorch LSTM for price-direction forecasts) with fundamental news analysis (running real-time sentiment extraction via VADER) — synthesizing all signals into a final recommendation.
Python · PyTorch · Claude APIFinancial Intelligence Assistant
We built a fully functional Financial Intelligence Assistant! It uses a collaborative multi-agent environment where specialized AI agents (like Technical/Price and Sentiment) work together to query live Alpha Vantage data and a local S&P 500 database.
PythonMachine Learning Chatbot
A Jupyter Notebook-driven chatbot experiment that dives deep into applied machine learning techniques. Built to test different approaches to natural language question answering and embedding representations.
Jupyter NotebookEmbedding Benchmarking
A robust benchmarking suite evaluating different text embedding models. Compares the performance, latency, and memory utilization of various open-source models tailored for retrieval-augmented generation (RAG).
Jupyter NotebookInteractive Web Apps
UW Foster MBA — AI Course Pilot
A RAG-powered course planning assistant built for UW Foster MBA students. Recommends electives based on career goals by searching 49 real course syllabi. Due to intellectual property restrictions, full access is limited to enrolled Foster MBA students.
Streamlit · OpenAI · RAG · Python
AI Stock Analysis Agent
Chat directly with an AI agent powered by Claude. Ask any stock question and the agent seamlessly merges complex technical stock analysis (training a custom PyTorch LSTM for price forecasts) with real-time news sentiment analysis (using VADER) — then delivers a synthesized BUY / HOLD / SELL recommendation.
Streamlit · Claude API · PyTorch
Financial Intelligence Assistant
We wrapped our Financial Intelligence AI system into a sleek Streamlit app! You can directly chat with our specialized agents to get live financial insights, or seamlessly swap between Baseline and Multi-Agent architectures.
Streamlit
Multi-Agent Chatbot Interface
A frontend wrapper for a powerful multi-agent architecture. Allows users to securely interact with specialized AI agents capable of maintaining persistent chat history and community moderation.
Streamlit
Embedding Benchmarking Tool
A visual analytics dashboard for comparing the performance, latency, and memory utilization of various open-source text embedding models tailored for RAG.
Streamlit
ML Chatbot (Mini Project)
An interactive machine learning chatbot built to demonstrate applied natural language processing techniques and custom conversational workflows.
Streamlit