Hey everyone! 👋 One of the biggest leaks in almost every local service business, clinic, or agency isn't client acquisition—it's speed to lead. When a prospect calls after hours or during a busy lunch rush and hits voicemail, over 70% won't leave a message. They just move on to the next business on Google. Alongside our work over at Oqvera, I’m expanding my skills and services directly to the freelance space on Fiverr to help more teams implement custom voice agents and workflow automation. Here is the exact breakdown of how this production voice architecture works under the hood so you can build or implement this for your own clients: 🛠️ The Tech Stack - Voice Engine & Telephony: Vapi AI / Retell AI + Twilio (SIP Routing) - Speech-to-Text & TTS: Deepgram Nova-2 + Cartesia / ElevenLabs - Brain / Reasoning: GPT-4o / Claude 3.5 Sonnet (with strict function calling) - Backend & Workflow Automation: n8n / Make.com + FastAPI webhooks - CRM & Scheduling: GoHighLevel / HubSpot + Google Calendar / Cal.com ⚙️ How the Workflow Operates 1. Low-Latency Inbound Handling: When a call hits the Twilio number, it routes to Vapi/Retell. Using Deepgram and Cartesia keeps the latency under 800ms, eliminating that robotic 2-second pause that makes callers hang up. 2. Knowledge Base & Guardrails: The system prompt is loaded with the business’s services, pricing guidelines, location, and operating hours. If a question is outside the scope, it gracefully deflects or offers to take a message rather than hallucinating. 3. Live Calendar Tool-Calling: When a caller says "I'd like to book a slot for Thursday afternoon," the LLM triggers a custom tool that pings the Google Calendar/Cal.com API, checks available slots in real time, and proposes exact times. Once agreed, it books the appointment instantly. 4. Post-Call CRM & SMS Pipeline (via n8n): The moment the call ends, an n8n webhook triggers: - Pushes the recording, full transcript, and structured summary to the CRM (GoHighLevel/HubSpot). - Fires an automated SMS confirmation to the caller with the appointment link and address. - If the caller mentioned an emergency keyword, it triggers a live warm transfer to an on-call manager.

Hey everyone 👋 I’m Saad, an AI/ML Engineer and developer, and today I’m excited to introduce something I’ve been building: OQVERA. I didn’t want to create another website that simply says: “We build AI solutions.” I wanted to build something that shows what I can actually build. So OQVERA is becoming my public engineering portfolio — a place where I showcase real projects, AI systems, automation workflows, agentic applications, and the technical work behind them. What you’ll find inside OQVERA 👇 🤖 AI & Agentic Systems Multi-agent architectures, AI agents, RAG systems, LangGraph workflows, intelligent assistants, and LLM-powered applications. ⚙️ Automation & Workflows Business process automation, API integrations, AI-powered workflows, calling agents, and systems designed to remove repetitive manual work. 💻 AI Engineering & Software Backend systems, APIs, AI integrations, computer vision, intelligent applications, and production-focused engineering. 🔬 Case Studies I’m documenting projects not just by showing the final UI, but by explaining the problem → architecture → technology → implementation → outcome. Why I built OQVERA I want my work to speak for itself. Instead of telling potential clients: “I can build an AI agent.” I can show them one. Instead of saying: “I know automation.” I can show the workflow. Instead of listing 20 technologies on a CV... I can show what I built with them. That’s the direction I want to take with OQVERA. And yes — I’m building this with clients in mind. If you're a founder, business owner, startup, or team looking at ways to use: - AI agents - LLMs - workflow automation - intelligent customer support - AI calling systems - RAG / knowledge systems - custom AI software - internal business automation I'd love for you to check out what I'm building. 🌐 OQVERA: https://oqvera.netlify.app/ This is only the beginning. I’ll be continuously adding new projects, experiments, architectures, and case studies as I build.

Hey everyone! 👋 ​I wanted to share a complete real-world Voice AI build I just wrapped up: Alex, an automated AI Voice Operations Receptionist for a pre-owned auto dealership (National Motors Inc.). ​If you’re building voice agents for local businesses, lead generation, or appointment setting, here is how the whole stack works under the hood. ​🔥 What the Voice Agent Handles Live on Calls: ​Real-time Inventory Lookups: Queries the live inventory database (make, model, mileage, pricing, stock numbers) and answers caller questions instantly. ​Test Drive Booking: Captures caller details (Name, Phone, Email, Preferred Date/Time) and logs them directly to a Google Sheet. ​Instant Email Confirmations: Automatically triggers HTML booking confirmation emails via Gmail right after hanging up. ​Trade-in & Financing Routing: Collects trade-in specs and guides callers through pre-qualification flows. ​🛠️ The Tech Stack: ​Vapi.ai: Voice Engine (STT/TTS/LLM orchestration with custom system prompts). ​Make.com: Central Webhook Router handling function tool execution. ​Google Sheets: Functioning as both a live database for dynamic lookups and a CRM lead sheet. ​Gmail API: Instant email confirmations. ​⚡ Key Takeaways & Common Pitfalls to Avoid: ​If you’re trying to build something similar in Vapi + Make.com, keep these technical details in mind: ​Array Indexing Matters: Vapi passes tool calls as a nested array (toolCalls[1]). If you map fields as toolCalls[] inside Make.com, your JSON strings will break, resulting in "No result returned" timeout errors! ​Routing Function Calls: Always set explicit Filter Rules directly on your Make.com Router modules (e.g., function.name EQUALS book_test_drive) so different tool execution branches don't trigger at the same time. ​Return Clean Webhook Responses: Always return an explicit 200 OK status with Content-Type: application/json header and a results array containing the toolCallId. Otherwise, the AI assistant will time out waiting for backend confirmation!

Hey everyone! 👋 Following up on my last update about thread persistence using SqliteSaver, I’ve upgraded the chatbot into a 11-tool agentic engine running on Groq (Llama 3.3 70B) and Streamlit. If you’ve ever tried binding 10+ custom tools to open-weight LLMs, you’ve probably hit APIError: Failed to call a function or malformed argument streams. Here is how I solved those issues and stabilized the loop: 🛠️ The 11-Tool Ecosystem - Live APIs & Web: get_weather (OpenWeatherMap), maps_tool (OpenStreetMap), web_search (DuckDuckGo) - Finance & Computation: get_stock_price, calculator, unit_converter - Productivity: send_email, calendar_event_manager, get_current_datetime - Context & Memory: personal_memory_store, search_conversation_history 💡 Key Technical Takeaways 1. Strict Pydantic Hardening: Instead of relying on raw docstrings or loose dict types, I attached explicit pydantic.BaseModel schemas to every @tool decorator. This forces Llama 3.3 to emit 100% deterministic JSON arguments. 2. Decoupled Execution: Separated function invocation from raw token streaming inside Streamlit to prevent stream-parsing collisions during tool calls. 3. Clean Graph State: Kept SQLite state checkpoints isolated so failed execution attempts don't contaminate thread history. 🧪 Multi-Tool Test Prompt I pushed the agent with a compound request: "Check current time ➔ get Tokyo weather ➔ convert °C to °F ➔ look up map coordinates ➔ store a note in memory." All tools executed sequentially in a single turn with live status blocks in the Streamlit UI! 🔗 Try the Live App: https://saad7420-chatbot-based-on-langgraph-app-utzuat.streamlit.app/ What tool-calling setups are you currently building with LangGraph? Drop your questions or code setup below and I'd be happy to share backend snippets! 👇

Hey everyone! I'm Muhammad Saad, an AI Engineer focused on building production-ready AI solutions. My expertise includes: • AI Agents & Multi-Agent Systems • LangGraph & CrewAI • RAG (Retrieval-Augmented Generation) • LLM Applications • AI Chatbots & Voice AI • Workflow Automation • FastAPI, Python, React & Next.js • API Integrations & Business Automation I've built AI systems for chatbots, automation workflows, research assistants, and custom business solutions. I enjoy taking ideas from concept to deployment. I'm looking to collaborate with founders, marketers, sales professionals, agencies, and other developers who want to build real AI products or start an AI company together. If you can bring ideas, clients, or business expertise, I can handle the technical side and help turn those ideas into reality. If you're serious about building something meaningful, send me a message. Let's create something great together. 🚀