AI Engineer · Automation Expert · RAG Systems · LLM Integration · Intelligent Workflows · Generative AI Solutions
AI Automation Engineer building AI-powered solutions that use Large Language Models to streamline financial advisory and insolvency processes. I design intelligent chatbot systems, implement RAG pipelines, and build scalable backend architectures, and I specialize in exposing ML/DL models through RESTful APIs using Flask and FastAPI.
I turn slow, manual workflows into fully automated, end-to-end systems — engineering data pipelines that pull from live public and third-party sources, orchestrating multi-agent and agentic tool-calling workflows, and shipping production AI on Docker and Azure. Strong foundation across LLMs, RAG, MLOps, computer vision, and workflow automation (Make.com, Monday.com, MCP).
- 🔭 Currently: Senior Technical Manager @ The Insolvency Group
- 🌍 Islamabad, Pakistan · Open to Work (On-site · Hybrid · Remote)
- 💬 Ask me about LLMs, RAG, AI agents, voice AI, and workflow automation
- Senior Technical Manager — The Insolvency Group · Oct 2024 – Present (previously Manager, Jun 2024 – Aug 2025) · United States · Remote
- Manager — Product & AI — BlueGrid.io · Aug 2020 – 2024 (progressed from AI Automation Engineer → Technical Team Lead → Manager) · United Kingdom · Remote
- Research Assistant — FAST-NUCES, Islamabad Campus · Aug 2019 – Jun 2020 · medical imaging research on the CheXpert chest X-ray dataset
AI & GenAI Development
Automation & Workflows
-FF3D57?style=flat&logo=mondaydotcom&logoColor=white){kind=icon}
Frameworks & Vector DBs
Voice & Realtime AI
-00B8D9?style=flat&logoColor=white){kind=icon}
Backend, Cloud & MLOps
-0078D4?style=flat&logo=microsoftazure&logoColor=white){kind=icon}
ML / DL & Foundations
| Project | Description | Tech Stack |
|---|---|---|
| AI Voice Call Agent | Production-grade realtime voice agent that handles live phone conversations end-to-end via a low-latency STT → LLM → TTS pipeline with natural turn-taking, interruption handling, and agentic tool/function calling for live retrieval during the call. | Python, LiveKit (Realtime Audio/WebRTC), Azure AI Foundry, Cartesia.ai, ElevenLabs, OpenAI/Anthropic LLMs, LangGraph, FastAPI, Docker |
| Multi-Agent Orchestrator | Framework that coordinates multiple specialized LLM agents to plan and execute complex, multi-step tasks, with tool/function calling, shared state, and routing between agents. | Python, LLM APIs, Agentic Orchestration, LangGraph, Tool/Function Calling |
| Agentic Document Research (RAG) | Document-grounded research agent that retrieves, reasons over, and synthesizes answers across large document collections using RAG and embedding-based semantic search — a domain-agnostic pattern for financial, legal, and insurance knowledge bases. | Python, LangChain, Vector Database, Embeddings, RAG, FastAPI |
| MCP Agent Toolkit | Reusable toolkit for building Model Context Protocol (MCP) agents that connect LLMs to external tools, APIs, and data sources through a standardized interface. | Python, Model Context Protocol (MCP), LLM APIs, Tool Integration |
| Medical AI Chatbot (RAG-Based) | Medical-domain chatbot using RAG and semantic search over clinical documents to deliver accurate, context-aware responses; deployed as a production AI service with a REST API and web interface. | Python, LLM APIs, LangChain, Vector Database, Embeddings, FastAPI, Streamlit, Docker |
| AI Personal Assistant | LLM-powered agentic assistant that understands intent, maintains conversational context, and autonomously executes multi-step tasks through tool/function calling and workflow orchestration. | Python, LangChain, Prompt Engineering, FastAPI/Flask, REST APIs, Docker |
| Audio-to-Text Application | End-to-end speech-to-text pipeline with automated recognition, speaker-aware conversation handling, transcript normalization, and summarization, built for both realtime and batch workflows. | Python, OpenAI Whisper / Speech Models, FastAPI, NLP, Audio Processing |
| Workflow Automation — Make.com & Monday.com | Production automation pipelines: a Make.com cart-review flow (webhook → HTTP → router → iterator → filter → Google Sheets) and a three-board Monday.com client-delivery system (Clients → Deliverables → Tasks) with GraphQL and idempotent upserts. | Make.com, Monday.com GraphQL API, TypeScript, JavaScript, Webhooks |
More projects → github.com/Mahnooramjad05 — CRM customer-service agent, computer-vision quality assurance, AI system-design patterns, AI paraphrasing tool.
- National University of Computer and Emerging Sciences (FAST) — MS Artificial Intelligence · 2020 – 2022
- National University of Computer and Emerging Sciences (FAST) — BS Computer Science · 2016 – 2020
- Deep Neural Network Bootcamp — GIK Institute, Swabi · Jun – Aug 2022
- View-Aware Design for Efficient Deep Learning in Chest X-Ray Interpretation — view-aware evaluation and deployment strategy for chest X-ray AI; lightweight CNNs reach near-DenseNet performance on frontal views at up to 3–4× lower inference latency, validated on PadChest with Grad-CAM interpretability.
- Graph-Aware Multi-View Learning for Chest X-Ray Diagnosis — modeling inter-pathology dependencies and cross-view complementarity on CheXpert; a Multi-View + GNN framework achieving the best overall AUC 0.7382.
- Ab initio insight into the physical properties of MgXH₃ (X = Co, Cu, Ni) lead-free perovskite for hydrogen storage application — Journal of Physics and Chemistry of Solids, 2023.
- A DFT study of structural, electronic, mechanical, phonon, thermodynamic, and H₂ storage properties of lead-free perovskite hydride MgXH₃ (X = Cr, Fe, Mn)
- Hydrogen Storage Capacity of Lead-Free Perovskite NaMTH₃ (MT = Sc, Ti, V): A DFT Study — Hindawi (Wiley Online Library).
💡 Open to AI Engineer / AI Automation roles and collaborations. Let's build systems that ship.