• Student at the University of Chicago pursuing a BS in Computer Science and a BA in Physics (expected 2028).
• Focused on building scalable, production-grade software and deployable AI/ML systems, with a strong foundation in systems, math, and engineering ownership.
• Actively seeking Software Engineering or AI/ML Engineering internship roles for Summer 2026.

These projects showcase demonstrated technical ownership, architecture, and real-world impact.

• AI-Driven Research Matchmaking Platform: A scalable, serverless application that intelligently matches students to research labs using semantic search and LLM-based compatibility scoring to replace the cold-email process.
• Production Architecture (GCP): Architected the entire solution on Google Cloud Platform (GCP) using Cloud Run and Cloud Load Balancing, orchestrated with Terraform (IaC) for reproducible deployment.
• Technical Stack: Built a modern full-stack application with Next.js 15 (React 19) and a high-concurrency FastAPI backend, utilizing ChromaDB for vector storage and Firestore for data persistence.
• Impact: Currently in pilot at the University of Chicago; designed for scaling to address academic communication and efficiency across higher education.

• Core System: Developed a modular Mixture-of-Experts (MoE) architecture for autonomous driving in the CARLA simulator, utilizing specialized expert networks and a gating network for decision-making. Built with PyTorch (DDP), CUDA, and Linux.
• Data Contribution: The project's pipeline resulted in two large-scale, public datasets for the autonomous driving research community:
  • CARLA Autopilot Multimodal Dataset (~365 GB, 82k frames): Synchronized RGB, semantic segmentation, LiDAR, 2D boxes, and ego-vehicle states.
  • CARLA Autopilot Images Dataset (~188 GB, 68k frames): Multi-camera images, control signals, and kinematics.
• Status & Learnings: Currently paused. The process provided deep expertise in high-performance data pipelines, distributed training, and the challenges of deploying AI systems.

• Ownership & Impact: Drove reliability and scalability improvements by diagnosing and fixing critical backend failures and executing major schema refactoring with zero downtime.
• Scalable Systems Design: Designed and deployed a queue-driven execution processing system to decouple heavy telemetry operations from the API, significantly reducing request latency and enabling horizontal scaling.
• Full-Stack Development: Developed full-stack admin analytics dashboards (NestJS, Next.js, Recharts) with SQL time-bucket aggregation, providing actionable insights into user growth and execution volume.
• Reliable Data Flow: Implemented a fault-tolerant, SQS-based background worker for telemetry aggregation and HubSpot CRM synchronization, ensuring reliable data delivery for downstream analytics and sales pipelines.
• Metrics & Observability: Introduced circuit execution metrics (complexity, duration) and a KPI dashboard for UTM-based marketing attribution, directly supporting growth strategy.

• Qwen vLLM on GKE: Cloud-native deployment pipeline for serving Qwen models on GKE Autopilot, provisioning NVIDIA T4 GPUs and deploying vLLM for a high-throughput, scalable inference endpoint.
• LocalRAG: Terminal LLM chat with infinite memory via FAISS-powered local vector search, enabling persistent, context-aware conversations without external servers.
• Semantic Image Search: Full-stack text-to-image retrieval: FastAPI backend, CLIP embeddings, and Next.js/Tailwind frontend.

• GovHub: A civic software concept offering a GitHub-style workflow for legislation. Built with React, Next.js, and TypeScript.
• Portfolio (ipeter.dev): This site, featuring ImmanuelAI—an LLM assistant (represented by biography.js) designed to interactively answer technical questions for recruiters.
• AI Commit: A Bash utility using the OpenAI API to automatically generate meaningful commit messages from staged diffs, improving engineering workflow quality.

• Digital Portfolio: ipeter.dev (Includes resume and ImmanuelAI chat)
• Hugging Face Datasets: huggingface.co/immanuelpeter