.png&w=640&q=75)
.png&w=640&q=75)
🤓 Latest Submissions
Tendril
Tendril is a live go-to-market (GTM) intelligence platform that converts scattered public web data, and spoken conversations, into evidence-backed, scored sales signals for B2B revenue teams. The core is a durable, multi-stage pipeline (discover, scrape, extract, graph, score, brief, outreach), where every stage commits state before yielding, so scans survive restarts. Bright Data powers live web acquisition: the SERP API runs targeted per-account queries, the Web Unlocker fetches bot-protected pages, and a Scraping Browser handles JavaScript-heavy fallbacks. AI/ML API acts as an OpenAI-compatible model gateway with routing: a cheap model for strict-JSON signal extraction, a stronger model (GPT-4o) for briefs, and a fast model for drafts. A parallel multimodal branch is the differentiator. Bright Data discovers public spoken sources, Tendril extracts the audio (yt-dlp and ffmpeg), and Speechmatics transcribes it live with speaker diarization and word-level timestamps. A cheap Featherless relevance filter gates the expensive extraction, and content-addressable hashing (SHA-256 over audio) guarantees the same media is never transcribed, or paid for, twice. Both branches converge into Cognee Cloud, a per-account knowledge graph that is written to and recalled from, grounding each brief in how the account changes over time. Engineering for trust and reliability is first-class. A transparent 0 to 100 rubric (fit, timing, relationship, evidence) gates "sales-ready", PII is scrubbed before any memory write, and per-scan budget caps plus a full provider-call audit trail keep cost and behavior accountable. Mock, live, and cached modes plus graceful fallbacks (Cognee to local, AI/ML to deterministic) mean a demo never breaks. Outreach is human-in-the-loop with ethical guardrails and switchable tone, never auto-sent. Tendril tells GTM teams who is ready now, why it matters, and what to say next, with every claim tied to a live, verifiable source.
31 May 2026
Vespergrid
VesperGrid helps industrial teams detect, understand, and respond to hazards before they escalate. It brings together evidence from cameras, drones, gas sensors, wind readings, voice reports, and operator notes, then turns that information into a clear incident view with affected zones, uncertainty, source evidence, and recommended response actions. Because real industrial hazard data is sensitive and difficult to access, this project uses a fully synthetic LNG terminal scenario. A Gazebo and ROS2 simulation generates the operational environment, including CCTV views, a drone feed, a visible gas plume, gas concentration changes, and wind drift. These simulated signals are sent into a FastAPI backend through an evidence ingest pipeline, where each input is processed and linked to the incident state. The system is designed for multimodal analysis on AMD MI300X. Visual evidence is parsed with Qwen2.5-VL served through vLLM, gas and wind traces are evaluated with deterministic safety logic for stable and auditable hazard scoring, and voice reports are transcribed with faster-whisper using a configured Whisper speech-to-text model. The processed evidence then flows into VesperGrid’s main orchestration layer, which combines all inputs into one source-linked operational state. From there, VesperGrid suggests possible response actions to the human operator, explains the evidence behind each action, highlights uncertainty, and shows the likely consequences of different choices before any action is approved. The final output is shown in a React command dashboard where operators can review live feeds, inspect evidence, understand risk zones, and initiate the next response. VesperGrid does not replace the human decision-maker. It gives operators a faster, clearer, and more accountable way to act when safety depends on minutes.
10 May 2026
