Computer Science + Physics at the University of Chicago. Incoming Member of Technical Staff Intern at Tensormesh.
- Portfolio: ipeter.dev
- Resume: ipeter.dev/resume.pdf
- LinkedIn: linkedin.com/in/immanuel-peter
- GitHub: github.com/immanuel-peter
- Hugging Face: huggingface.co/immanuelpeter
- Email:
ipeter@uchicago.edu
- Joining Tensormesh to work alongside the team behind LMCache on inference infrastructure.
- Interested in software engineering, developer infrastructure, and self-driving neural networks.
Hostess is a deployment platform for multi-service applications built around a declarative hostess.yml. It combines a Go CLI, control plane, and Studio dashboard to deploy full stacks like Next.js, FastAPI, Postgres, and Redis with generated Kubernetes manifests, service discovery, secrets wiring, per-service deploys, and framework-specific operational views.
- Live: hostess.sh
- Stack: Go, Kubernetes, Next.js, PostgreSQL, GCP, Docker
A Kubernetes Redis operator inspired by CloudNativePG's control-plane design. It manages pods and PVCs directly instead of relying on StatefulSets, which lets it enforce fencing-first failover, replica-first rolling updates, pod-level instance management, backup workflows, and deterministic behavior across standalone, sentinel, and cluster modes.
- Source: howl-cloud/redis-operator
- Stack: Go, Kubernetes, Redis, Helm, Prometheus
A modular self-driving research stack built around a Mixture-of-Experts architecture instead of a single end-to-end model. The repo includes data pipelines for BDD100K, nuScenes, and CARLA, specialized perception experts, a context-aware gating network, a trajectory policy head, and released CARLA datasets on Hugging Face. The final integrated simulation stage is paused, but the project captures the full training, evaluation, and research workflow.
- Write-up: Building AutoMoE
- Source: immanuel-peter/self-driving-model
- Stack: Python, PyTorch, DDP, CUDA, CARLA, Hugging Face
An AI-powered research matching platform that connects students with labs using semantic search and LLM-based fit scoring. It replaces fragmented outreach with a centralized pipeline for discovery, parsing, ranking, and review.
- Write-up: Inside the Architecture of Matchbox
- Stack: Next.js, FastAPI, ChromaDB, GCP, OpenAI API
A small product that turns a public GitHub pull request URL into a streaming AI code review with line-by-line feedback.
- Live: grokreq.com
- Source: immanuel-peter/grok-review
- Stack: Next.js, TypeScript, GitHub API, xAI API
Incoming Member of Technical Staff Intern. Tensormesh helps enterprises reduce GPU cost by offloading reusable KV caches during inference.
Software Engineer Intern. Delivered 19 PRs and 43 contributions across schema refactors, queue-driven execution processing, telemetry aggregation with AWS SQS + TypeORM, and full-stack analytics dashboards with NestJS, Next.js, and Recharts.