Tooba M.

Your team is probably spending far more than it should on work that AI can already handle. I build production-grade AI systems that automate that work, reduce operational costs, and give your team back hours every week. Over the past 12 years, I’ve built and shipped 25+ production AI systems across healthcare, life sciences, fintech, distribution, hospitality, construction, and enterprise SaaS. 𝐖𝐡𝐚𝐭 𝐈 𝐛𝐮𝐢𝐥𝐝 𝐀𝐈 𝐀𝐠𝐞𝐧𝐭𝐬 & 𝐀𝐮𝐭𝐨𝐦𝐚𝐭𝐢𝐨𝐧 Multi-agent AI systems Autonomous workflow automation AI copilots and internal assistants MCP-powered business agents Human-in-the-loop workflows AI integrations with existing business systems 📄 𝐃𝐨𝐜𝐮𝐦𝐞𝐧𝐭 & 𝐃𝐚𝐭𝐚 𝐈𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞 Intelligent document processing OCR and PDF extraction Structured data extraction Financial and regulatory document analysis Automated report generation RAG and enterprise knowledge systems 🎙 𝐕𝐨𝐢𝐜𝐞 & 𝐂𝐨𝐧𝐯𝐞𝐫𝐬𝐚𝐭𝐢𝐨𝐧𝐚𝐥 𝐀𝐈 AI voice agents AI receptionists Appointment and reservation agents WhatsApp automation Customer support automation OpenAI Realtime, Retell AI, ElevenLabs 🔎 𝐄𝐧𝐭𝐞𝐫𝐩𝐫𝐢𝐬𝐞 𝐒𝐞𝐚𝐫𝐜𝐡 & 𝐊𝐧𝐨𝐰𝐥𝐞𝐝𝐠𝐞 𝐒𝐲𝐬𝐭𝐞𝐦𝐬 RAG platforms Internal knowledge assistants Semantic and hybrid search Computer vision / reverse image search Connecting AI to ERP, CRM, email, tickets, and company knowledge A few systems I've built 🧬 𝐋𝐢𝐟𝐞 𝐒𝐜𝐢𝐞𝐧𝐜𝐞𝐬 𝐑𝐞𝐠𝐮𝐥𝐚𝐭𝐨𝐫𝐲 𝐀𝐈 Built a multi-agent platform that automates IND/CTA submissions, clinical study reports, and safety narratives for biotech and pharmaceutical organizations. The system extracts information from documents, generates regulatory templates, propagates data across modules, and includes human validation workflows. Stack: Google ADK, LiteLLM, AWS, Agentic AI Compliance: SOC 2, GDPR 💰 𝐅𝐢𝐧𝐚𝐧𝐜𝐢𝐚𝐥 𝐃𝐨𝐜𝐮𝐦𝐞𝐧𝐭 𝐈𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞 Built a Claude-powered pipeline for a New York investment firm that processes 200+ page financial documents, classifies content, extracts structured information into a governed database, and generates defensible outputs. The system also includes an evaluation framework and edge-case dataset to continuously measure accuracy. Stack: Claude, Document AI, Structured Extraction, LLM Evals 𝐀𝐈 𝐊𝐧𝐨𝐰𝐥𝐞𝐝𝐠𝐞 + 𝐕𝐢𝐬𝐮𝐚𝐥 𝐒𝐞𝐚𝐫𝐜𝐡 𝐟𝐨𝐫 𝐉𝐞𝐰𝐞𝐥𝐫𝐲 𝐌𝐚𝐧𝐮𝐟𝐚𝐜𝐭𝐮𝐫𝐢𝐧𝐠 Built an AI knowledge assistant that connects Microsoft Teams with emails, Zoho tickets, and a legacy NAV 2009 ERP. Also built a custom computer-vision search engine allowing B2B customers to find jewelry products using images. Stack: Azure OpenAI, RAG, FastAPI, Computer Vision 📦 𝐀𝐈 𝐎𝐫𝐝𝐞𝐫 𝐏𝐫𝐨𝐜𝐞𝐬𝐬𝐢𝐧𝐠 𝐟𝐨𝐫 𝐚 𝐃𝐢𝐬𝐭𝐫𝐢𝐛𝐮𝐭𝐨𝐫 Automated purchase-order processing across 25,000+ SKUs, reducing processing time from approximately 30 minutes to under 1 minute with 97% accuracy. Stack: Vertex AI, Gemini, Claude, MCP, NetSuite 𝐕𝐨𝐢𝐜𝐞 + 𝐖𝐡𝐚𝐭𝐬𝐀𝐩𝐩 𝐀𝐈 𝐟𝐨𝐫 𝐇𝐨𝐬𝐩𝐢𝐭𝐚𝐥𝐢𝐭𝐲 Built a voice and WhatsApp automation system for a multi-venue restaurant group that handles 95%+ of reservations and event inquiries, generating approximately £30K in annual savings while preventing missed leads. 🩺 𝐇𝐞𝐚𝐥𝐭𝐡𝐜𝐚𝐫𝐞 𝐀𝐈 𝐀𝐬𝐬𝐢𝐬𝐭𝐚𝐧𝐭 Built a voice, chat, and OCR assistant for a major diagnostic laboratory, supporting test discovery, appointment booking, prescription parsing, symptom-related interactions, and report interpretation. 𝐌𝐲 𝐭𝐞𝐜𝐡𝐧𝐢𝐜𝐚𝐥 𝐬𝐭𝐚𝐜𝐤 AI & Agents: OpenAI, Claude, Gemini, AWS Bedrock, LangChain, LangGraph, LlamaIndex, Pydantic AI, Google ADK, MCP Voice AI: OpenAI Realtime API, Retell AI, ElevenLabs, Whisper RAG & Document AI: Pinecone, Qdrant, Weaviate, ChromaDB, OCR, PDF extraction, document intelligence Backend: Python, FastAPI, Django, PostgreSQL Frontend: React, Next.js Cloud & Infrastructure: AWS, Azure, GCP, Docker Integrations: NetSuite, Salesforce, HubSpot, Slack, Stripe, QuickBooks, Zoho, Microsoft Teams, Zapier, Make AI Observability & Evaluation: LangSmith, Logfire, Arize Phoenix, custom LLM evaluation frameworks 𝐇𝐨𝐰 𝐈 𝐰𝐨𝐫𝐤 I don't start by throwing an LLM at your problem. First, I understand the workflow, identify where automation will actually create business value, define measurable success criteria, and then design the system around those requirements. You get: → Clear scope before development starts → Defined milestones and deliverables → Production-ready architecture → Human-in-the-loop safeguards where needed → Monitoring and evaluation → Integration with your existing tools → Documentation and handover → End-to-end ownership from idea to production I've maintained a 93% on-time, on-budget delivery rate and typically respond within 30 minutes. 𝐈'𝐥𝐥 𝐡𝐞𝐥𝐩 𝐈'𝐥𝐥 𝐡𝐞𝐥𝐩 𝐲𝐨𝐮 𝐝𝐞𝐭𝐞𝐫𝐦𝐢𝐧𝐞 𝐰𝐡𝐚𝐭 𝐬𝐡𝐨𝐮𝐥𝐝 𝐛𝐞 𝐚𝐮𝐭𝐨𝐦𝐚𝐭𝐞𝐝, 𝐰𝐡𝐚𝐭 𝐬𝐡𝐨𝐮𝐥𝐝𝐧'𝐭, 𝐚𝐧𝐝 𝐰𝐡𝐚𝐭 𝐚 𝐩𝐫𝐚𝐜𝐭𝐢𝐜𝐚𝐥 𝐩𝐫𝐨𝐝𝐮𝐜𝐭𝐢𝐨𝐧 𝐬𝐨𝐥𝐮𝐭𝐢𝐨𝐧 𝐰𝐨𝐮𝐥𝐝 𝐥𝐨𝐨𝐤 𝐥𝐢𝐤e