Muhammad
Ahmed
AI/ML Engineer & AI Developer
Agentic AI • Voice Systems • LLM Fine-Tuning • Workflow Automation
I build production AI systems across agentic workflows, voice AI, LLM fine-tuning, and business automation. Currently building AI systems and governance at JETZT.
I turn language models into reliable, production-grade systems — orchestrating agents, grounding them in real data, and shipping the automation around them.
LLM & Agents
ML & Vision
Data & Vectors
Engineering
Experience
AI Systems Engineer & AI Governance Lead
- Building production AI systems including a real-time voice AI phone agent, multi-agent pipelines, and AI-driven automation for real-world business use cases.
- Shaping internal AI governance and responsible-AI practices, guiding how AI capabilities are adopted and deployed across the organization.
AI/ML Engineer Intern — AI, Cybersecurity & Disaster Response
- Led the AI intern cohort building an AI-powered hybrid disaster alert system for Pakistan, covering LoRa mesh networks, SMS Cell Broadcast, and offline AI deployment with differentiated models for Punjab and KPK.
- Applied research at the intersection of AI, cybersecurity, and the UN SDGs, under the mentorship of Ammar Jaffri, founder of PISA and Chairman of Pakistan’s National Cyber Security Task Force.
AI/ML Research Collaborator — Medical Imaging & Energy Optimization
- Collaborated with PhD and PostDoc researchers on medical imaging systems, including skin cancer detection and autism subtype classification models.
- Built an energy consumption optimization system leveraging Agentic AI and ML models.
Projects
JETZT Voice Assistant
A production inbound voice AI agent for JETZT: Telnyx media streaming into Deepgram STT, a LangGraph/GPT-4o reasoning layer with live tool access, and Deepgram TTS streamed back to the caller sentence-by-sentence for low-latency conversation.
Multi-Agent Task Manager
A CrewAI system with hierarchical delegation: a Planner agent breaks a request down and routes work to dedicated Slack, Jira, and Memory agents, backed by Weaviate semantic memory, with retries, approval workflows, and both a FastAPI and Streamlit interface.
RickLLM
Fine-tuned LLaMA 3.1 8B on a Rick and Morty transcript corpus (converted to a ChatML instruction dataset) using QLoRA via Unsloth on a single Colab T4 GPU, then quantized to GGUF for offline inference through Ollama with a Streamlit chat UI.
Custom Cold-Calling Dialer
A full outbound-calling product: browser-based WebRTC calling and phone-bridge calling side by side, Telnyx-verified webhooks, call recording to Supabase Storage, CSV/XLS/XLSX lead ingestion, dispositions, callback scheduling, and a role-gated admin dashboard.
Rocket Task AI
LangGraph multi-agent system integrated with Rocket.Chat via webhooks. Four coordinated agents (Task Detector, Assignment Recommender, Employee Profiler, and Summarizer) with semantic vector search over Qdrant, JWT auth, and full Docker Compose deployment.