Team
Beyond Chat
Project Concept
No description has been added yet.
Entry
Status: Submitted
Last saved: May 09 at 5:58 PM EDT
Team Roster
Message board not available for this team yet.
Paramjeet Singh Team Lead RSVP Approved
Student at Northeastern University
Owned the LangGraph agent path and CopilotKit integration story: backend tool analyze_boxing_clip and related state (coaching_data, coaching_status, coaching_score_history) in apps/agent/src/agent.py; TwelveLabs integration in apps/agent/src/twelvelabs_client.py (Pegasus analysis, Marengo search, indexed video metadata / hls.video_url handling); environment and runbook setup (apps/agent/.env, COACHME_SKIP_NOTION, cache vs live COACHME_USE_CACHE). Worked with Gemini via LangChain (langchain-google-genai) and the CopilotKit BFF (LangGraphAgent, deployment URL). Optional: LangSmith tracing if configured. Git / branch ownership and submission materials (README, demo narrative).
I'm an AI/ML Engineer with 3+ years of production experience building LLM pipelines, RAG systems, and agentic workflows, most recently at Acuvate Software, where I shipped systems for large enterprise clients including a complaint classification system that cut manual processing by ~60%. I just completed my MS in Information Systems at Northeastern University ( GPA 3.72) and recently won the Grand Prize at the Voxel51 × TwelveLabs Video Understanding AI Hackathon with CoachMe+, an AI sports coaching tool powered by multimodal video understanding. I'm energized by early-stage environments where I can go from idea to deployed product fast which is exactly what I did building ConNET solo at the Cursor Hack-a-Sprint 2026
Agentic AI systems and multi-agent orchestration at scale. Multimodal and video understanding models especially applied to real-world domains like sports, health, and education. MCP and tool-augmented LLMs. Early-stage product development where AI is core to the product, not a feature. Always interested in connecting with builders working at the intersection of AI infrastructure and end-user applications.
Currently working on two projects: CoachMe+, an AI sports coaching tool built on top of FiftyOne and TwelveLabs' video understanding models, and ConNET, an agentic AI networking assistant with email, vault, and phone capabilities
Manisha Sahu RSVP Approved
Student at Northeastern University
Next.js /coach experience: apps/frontend/src/app/coach/page.tsx, CoachingCanvas, coaching components under apps/frontend/src/components/coaching/ (e.g. timeline, drills, session trend), and CopilotKit v2 sidebar UX. Styling in coaching-theme.css (layout, hero, cards). useFrontendTool flows in chat (e.g. drill cards, focus, comparison). Testing: vitest / pytest runs or test cases. Demo: 2‑minute script, recording, or form copy. Docs: README / dev-docs edits.
Frontend and AI application engineer focused on building interfaces where AI agents don't just return text — they render interactive UI. I build at the intersection of agentic systems and real-time frontend experiences.
Most recently built CoachMe+ at the Generative UI Global Hackathon — an AI boxing coach where the agent generates the entire training dashboard at runtime from video analysis. I owned the full frontend vertical slice: CopilotKit integration, generative UI components, agentic feedback loops (approve/skip flows that change what the agent recommends next), and a dark sports-tech design system with animated score rings and progressive rendering.
Previously contributed to the CoachMe+ video analysis pipeline that won the Voxel51 × TwelveLabs Hackathon. I care about making AI f
Product review, technical architecture, developer marketing, knowledge sharing with peers, community and friendships, design partners, AI engineering, generative UI, agentic systems, real-time frontend experiences, Model Context Protocol (MCP), LangGraph agent integration, full-stack development, machine learning, computer vision.Generative UI — how AI agents should render interactive interfaces instead of returning plain text. Interested in AG-UI protocol adoption, A2UI patterns, and MCP Apps a
Building CoachMe+ — an AI boxing coach that generates your entire training interface at runtime. The agent analyzes footage via an MCP server and decides what UI to render: animated score rings, drill cards with approve/skip feedback loops, correction timelines, and session summaries. Stack: Next.js, CopilotKit v2 (useAgent, useFrontendTool), LangGraph, AG-UI, A2UI, MCP Apps. Dark sports-tech theme, progressive staggered rendering, and three generative UI tools the agent calls mid-conversation.