Wajahat Ali B.

91% RAG retrieval accuracy. 34% LLM accuracy gain over the base model. 61% support workload reduction through AI agents. 6 production AI systems shipped.** I’m a Senior AI/ML Engineer with 4+ years of experience focused on building and deploying production AI systems from LLM fine-tuning and RAG to agentic workflows and multimodal AI. Most AI systems work in a demo. The hard part is making them reliable with real data, measurable evaluation, latency constraints, and real users. That’s where I focus: **architecting, benchmarking, optimizing, and deploying AI systems that hold up beyond the prototype.** ─────────────────────────────────────── WHO I WORK WITH ─────────────────────────────────────── → Founders building production AI products who need technical ownership, not just implementation → Startups moving from MVP to a reliable system used by real customers → Engineering teams integrating LLMs, RAG, and AI agents into existing products → Businesses looking to automate workflows, improve operational efficiency, or deploy custom AI models ─────────────────────────────────────── PROVEN RESULTS ─────────────────────────────────────── ✅ RAG Document Intelligence 91% retrieval accuracy across 10,000+ documents. Reduced query time from 35 minutes to under 2 seconds. ✅ LLM Fine-Tuning Improved LLaMA 3 task accuracy from 66% → 89% using parameter-efficient fine-tuning, while reducing hallucination rate by 71%. ✅ Multi-Agent Research System Reduced report generation time from 8 hours → 22 minutes using multi-agent orchestration and automated validation. ✅ AI Voice & Chat Agent Resolved 68% of Tier-1 support queries autonomously, reducing support workload by 61%. ✅ Computer Vision Pipeline Achieved 45 FPS on Jetson Nano with YOLO and TensorRT optimization, reaching 0.89 [email protected]. ✅ AI SaaS Platform Built and shipped a complete AI platform in 6 weeks, reducing document processing time from 4 minutes → 18 seconds with 97.3% extraction accuracy. ─────────────────────────────────────── WHAT I BUILD ─────────────────────────────────────── LLMs & Fine-Tuning LLaMA, Mistral, QLoRA, LoRA, PEFT, supervised fine-tuning, evaluation, inference optimization. RAG & Knowledge Systems Hybrid retrieval, vector search, reranking, document intelligence, structured extraction, and knowledge-base systems. AI Agents & Automation LangGraph, LangChain, CrewAI, multi-agent orchestration, tool use, workflow automation, and agent evaluation. Multimodal & Computer Vision VLMs, object detection, OCR, image understanding, and edge inference. Production AI & MLOps Python, FastAPI, Docker, Kubernetes, PyTorch, Hugging Face, vLLM, AWS, GCP, and RunPod. ─────────────────────────────────────── HOW I WORK ─────────────────────────────────────── I start with the problem and success metric, not the model. → Define the measurable objective → Establish a baseline → Design the architecture → Build and evaluate → Optimize accuracy, latency, and cost → Deploy and monitor in production No unnecessary AI. No demo-first engineering. The goal is a system that actually works. ─────────────────────────────────────── LET'S TALK ─────────────────────────────────────── Have an AI system you’re trying to build, scale, or fix? Send me the problem, current setup, and target outcome. I’ll review it and respond with whether I can help, the likely technical approach, and the key questions I’d investigate. Straight technical discussion. No sales pitch.