Our Work

Real systems in production — AI agents, knowledge platforms, and the infrastructure that keeps them reliable.

GovToKnow

RAG / GovTech

Gold-standard RAG chatbot for municipal government data

  • 3+ municipalities onboarded with 12K+ ordinance chunks indexed
  • 4 distinct chunking strategies (v1-v4) for different document types
  • 100+ adversarial tests validating RAG accuracy
  • Live at govtoknow.com serving real municipal queries
View Project

ChemMasters Data Platform

AI Data Platform

AI-powered data platform for a chemicals company

  • Crash-safe email pipeline using PG advisory locks for per-email mutual exclusion
  • Unified Email Agent (Claude) classifies, extracts, and routes inbound mail end-to-end
  • Companion Email Response Agent drafts replies for human review
  • Truth-set evaluation harness keyed by Sage OR# regression-tests every prompt change before promotion
View Project

A3E Environmental — New Flo CRM & AI Assistant

Enterprise AI + CRM

Enterprise AI assistant + full CRM for environmental consultants

  • Pioneered MCP server implementation with 20+ production tools
  • SparkyBot AI assistant handling daily operations via natural language
  • Full CRM with Gmail sync, proposals, and Stripe payments
  • Zero-downtime dynamic tool loading — add capabilities without restarts
View Project

GuruCloudAI Platform

SaaS Platform

AI agent orchestration and development automation SaaS

  • Full AI agent orchestration with multi-agent mesh coordination
  • Knowledge Bank with semantic vector search
  • Document Review system with audio feedback via Whisper
  • Dev VM management with auto-provisioning and session tracking
View Project

OfferEngine — Real Estate AI Platform

Full-Stack AI Platform

Automated real estate acquisition: MLS to contract in 2 hours

  • 2M+ properties analyzed with LLM-derived valuations
  • 2TB+ database with sub-second queries
  • Full automation: MLS listing to contract in 2 hours
  • 100% email deliverability for automated campaigns
View Project

Corporate Soul Starter

Multi-Agent RAG

Multi-agent FAQ/knowledge system for corporate environments

  • 5 specialized agents each with domain-specific RAG
  • S3→SQS→Textract pipeline for automated document ingestion
  • Session memory supporting 30 turns with 72-hour TTL
  • Kubernetes-ready with Helm charts
View Project

GuruCloud KB Python SDK

Developer Tool / SDK

Published PyPI SDK for Knowledge Bank API

  • Published on PyPI — pip install gurucloud-kb
  • Sync and async client support
  • Multi-dimensional weighted semantic search
  • MCP integration for AI agent workflows
View Project

Genie — Self-Evolving CRM

AI-Driven CRM

AI-driven CRM built autonomously by agents

  • AI agents autonomously implement user-requested features
  • Wish Fulfillment Pipeline from natural language to deployed code
  • AI assistant contextually embedded in every page
  • Data centralization ETL from scattered sources
View Project

DreamScape

AI + 3D / Creative

AI-powered interactive 3D world builder

  • Real-time 3D scene generation from natural language
  • Voice input via Web Speech API
  • GPT-4o function calling for object creation
  • ~$0.01-0.05 per interaction cost
View Project

Coltivera

Crowdfunding / AI Platform

AI-powered idea incubator with crowdfunded POC builds

  • All-or-nothing crowdfunding for POCs
  • AI agents build working demos autonomously
  • Revenue share tracking for ideators and investors
View Project

Cadenza — AI Grant Writing Platform

AI / Grant Writing

Self-learning AI platform for nonprofit grant writing

  • 75% autonomous error reduction via self-learning PI system
  • 80% reduction in report creation time
  • Multi-tenant architecture for 500+ concurrent organizations
  • 95% relevance accuracy in multi-stage retrieval
View Project

Mera / NuSeat — AI Job Matching Platform

AI / Job Matching

Dynamic LLM-powered job matching with L-Trees algorithm

  • Novel L-Trees algorithm for dynamic LLM-guided categorization
  • CL-Trees chronological variant for temporal memory
  • DynamicContextObjects abstraction (pgvector-based)
  • Overlap with key ideas in Microsoft's GraphRAG
View Project

With.in — Vendor-Side Auth Proxy for AI Agents

AI Infrastructure / Lead Gen

Lead generation through AI agent authentication workflows

  • 6-state FSM lifecycle for authentication flows
  • 4 authentication scenarios (no creds, vendor only, Within only, both)
  • SSE event stream for real-time visibility
  • Pyright strict mode — full type safety
View Project

Polymarket Speed Arbitrage Bot

Fintech / Trading

News-price correlation analysis and automated trading

  • 78.4% arbitrage opportunity score (EXCELLENT)
  • 185-minute median lag time identified
  • 1,124 correlations analyzed across 13 markets
  • Backtesting engine with multiple strategies
View Project

AI Dungeons & Dragons

Real-Time Voice / Multi-Agent AI

Voice-driven AI tabletop: an AI Dungeon Master and autonomous AI companions with real-time talking-head avatars

  • Voice-in / voice-out play loop in production: spoken player actions transcribed inline (Whisper), character lines rendered as emotion-directed speech (12 delivery cues) or live WebRTC talking heads
  • Fast-turn architecture: player-facing narration returns inline while authoritative game-state adjudication runs on a stronger model asynchronously — per-call latency_ms instrumented and recorded
  • Round-broker turn arbitration: companions propose lines in parallel with self-reported urgency, the winner's line feeds back so remaining companions react to it — bounded cost per beat, no crosstalk
  • Live avatars are strictly script-driven: the vendor's own LLM is locked out and input audio disabled so nothing but our DM can make a character speak
View Project

AI Avatar Client Meetings

Real-Time Voice / Conversational AI

A live, real-time conversational AI avatar that runs actual client meetings — in the browser or inside Zoom/Meet/Teams

  • In production for real prospect calls sourced from Upwork — meetings offered, booked, held, and followed up autonomously by an AI agent
  • Leg-chaining beats the vendor's 10-minute session cap: fresh sessions seeded with the transcript so far, persona resumes mid-thought, one continuous recording across handovers
  • Burn-free preflight mints a real session token from the real persona config — proving the stack end-to-end without spending meeting minutes
  • Invite path places the avatar inside the client's own meeting as a disclosed, badged AI participant
View Project