Team

.

Project Concept

No description has been added yet.

Entry

Status: Submitted

Last saved: May 09 at 6:02 PM PDT

View Entry

Team Roster

Message board not available for this team yet.

Aman Goyal Team Lead RSVP Approved

Agentic AI Product Manager at T-Mobile

I wrote the agent prompt and the demo data. The prompt tells the LLM to output three things: a flow diagram as nodes and edges, reviewer comments tied to ex act quotes from the text, and a spec card when the user requests it. We started with Gemini but switched to OpenAI because Gemini kept changing node IDs and paraphrasing quotes. I also wrote the four fixture files that define what the agent output should look like after each demo paragraph. I used Claude to help debug why the prompt was producing invalid JSON and to check the acceptance criteria logic. Getting stable node IDs across multiple agent runs was the harde st problem.

Aman Goyal is an Agentic AI Product Manager at T-Mobile. He is a Technical Product Manager and tinkerer, looking for knowledge sharing and co-founders. Aman is vegetarian and employed.

Caeden Kidd RSVP Approved

MIchigan Tech Student + Biomedical Machine Learning Intern at email me

I built the runtime middleware and state validation for the agent. I wrote the middleware that declares the Pair-PM fields on the agent state schema so Copi lotKit does not delete them during state snapshot round-trips. I replaced the old lead-triage middleware with the new Pair-PM middleware. I also wrote the v alidator, which parses and checks every agent output against the expected shape. If validation fails, it logs the error and returns the previous valid state instead of crashing the UI. I cleaned up the agent entry point to remove unused Notion and lead-store imports. I used Kimi to help scaffold the middleware s tructure and figure out where to hook the validator into the agent lifecycle.

I'm a senior CS student at Michigan Tech. I do ML research on medical imaging, specifically deep learning for coronary artery disease diagnosis. Outside the lab I'm a serial hackathon builder. Recent projects include an AI-powered language learning app, a Polymarket insider trading detector using tree-of-thought reasoning, and a clinical communication trainer for medical professionals. I've been building since I was 10, starting with my first website and working my way up to neural networks from scratch. I love problems that are technically hard and actually matter.

AI for healthcare and medical imaging, agentic systems, language learning and NLP. I want to connect with researchers, founders, and builders working on applied ML, especially clinical tools and real-time AI systems. I'm also always looking for hackathon teammates and mentors with experience taking technical projects from prototype to something people actually use.

I'm doing ML research at Michigan Tech building deep learning models in PyTorch for automated analysis of coronary angiograms, trying to give cardiologists more reliable quantitative tools for diagnosing heart disease. I also recently built Maizu, a language learning app that treats vocabulary as a connected graph instead of a linear sequence. Words link to other words and you learn them in clusters. On the side I keep building at hackathons: recent projects include an agentic LLM insider tradin

Ritunjay Murali RSVP Approved

AI Engineer Intern at Intellect Design Arena

I built the entire frontend. The layout is a 50/50 split: a text editor on the left, and a right column split 60/40 between a flow diagram and comment cards . I wrote the editor with a 2-second debounce and blur push to shared state. I wrote the diagram components, which render the agent's nodes and edges. I imp lemented stable-ID diffing so existing nodes stay in place and new nodes animate in with a 200ms staggered fade and a 400ms green flash. I wrote the comment cards with open and resolved states, a resolve button, and yellow node pulsing on hover. I wrote the spec card with the refined PRD, acceptance criteria, re solved questions, and a copy-to-clipboard button. I used Claude to help with the text editor integration and the custom diagram nodes.

MS CS graduate (UC Riverside) and AI/ML engineer with experience at a fintech AI platform GTM team and a nonprofit. I build LLM agents, RAG pipelines, and ML systems — from AML/KYC automation for Tier 1 banks to LiDAR-based canopy classification. Published IEEE researcher with a full-stack background.

Eager to explore: production agentic systems at scale, knowledge graphs for enterprise RAG, efficient on-device ML inference, and AI in fintech. Open to collaborating on open-source tooling for LLM observability, evaluation frameworks, or retrieval infrastructure.

Building production-grade agentic infrastructure: VectorScale, a self-hosted retrieval system (gRPC/FAISS/Kubernetes) cutting costs 95% vs managed alternatives; and CodeLens, an ML reranking pipeline for code review with 9× model compression and 5× inference speedup via LoRA + ONNX distillation.