Mirsaid A.

Senior AI Engineer | AI Agents, RAG, n8n Automation | Computer Vision

AI Engineer. Production AI agents, RAG, Voice AI and computer vision. Legal AI platform (wakil.ai) : 7,000+ users, 500+ paying subscribers. Retail vision: 90%+ accuracy, 80% less audit time. 🤖 AI AGENTS, RAG & AUTOMATION I design and build AI agents that can retrieve information, reason over data, use tools, call APIs, execute workflows, and safely hand off to humans when needed. One of my main projects is WakilAI, a production multilingual legal AI platform serving 7,000+ users and 500+ paying subscribers Check it out: wakil.ai I worked on its AI architecture and RAG development across ingestion, retrieval, reranking, agent routing, prompting, and evaluation. We improved answer accuracy from 38% → 51% and achieved 17% higher accuracy than competing legal assistants across 22+ evaluation categories and 4 languages. The system evolved beyond standard RAG into an agentic architecture with adaptive retrieval, domain routing, memory, retrieval quality gates, multilingual search, and GraphRAG capabilities. I also build practical business automation. One recent system combined n8n, Chatwoot, Qdrant, Whisper, and LLMs into a self-hosted AI support agent handling both voice and text. It performs knowledge retrieval, generates grounded responses, evaluates confidence, and automatically escalates uncertain conversations to human support. Agentic AI stack: LangGraph · LangChain · CrewAI · n8n · Tool/Function Calling · MCP · RAG · GraphRAG · Qdrant · Milvus · pgvector · Elasticsearch · Langfuse · RAGAS 👁️ COMPUTER VISION & EDGE AI Computer Vision is an equally important part of my background. I have built production systems for: - Object detection and classification - Multi-object tracking and ReID - Face recognition - Retail shelf and SKU analysis - Industrial inspection - OCR and equipment-tag extraction - Anomaly and condition detection - Real-time video analytics - NVIDIA Jetson / edge deployment / TensorRT I led development of a retail shelf-analysis system that achieved 90%+ product-detection accuracy and helped reduce manual shelf-audit time by 80%, with deployment across hundreds of retail locations. I also built an industrial field-intelligence platform combining Android image capture, YOLO, Grounding DINO, OCR, P&ID verification, asynchronous AI workers, a React operations dashboard, and a grounded AI Copilot - bringing Computer Vision and Agentic AI together in one production workflow. I’m also a co-author of a peer-reviewed IEEE Access paper on real-time Multi-Object Tracking, covering practical evaluation of detection, ReID, tracking, and edge-device performance. Computer Vision stack: PyTorch · OpenCV · YOLOv8–YOLO11 · Grounding DINO · ByteTrack · BoTSORT · DeepOCSORT · StrongSORT · InsightFace · ONNX · TensorRT · NVIDIA Jetson 🎙️ VOICE AI I’ve developed real-time voice AI systems for automated customer communication, including a native voice assistant with sub-3-second response times used across 6+ industries. I work with speech-to-text, LLM orchestration, streaming, WebSockets, voice workflows, and human escalation. ⚙️ PRODUCTION AI ENGINEERING Building the model is only part of the job. I can take AI systems through the full production lifecycle: Python · FastAPI · PostgreSQL · Redis · Celery · Docker · REST APIs · WebSockets/SSE · React/TypeScript · AWS/GCP/Azure · vector databases · monitoring · evaluation · deployment I have also built large-scale ingestion and search pipelines, including asynchronous document-processing systems using FastAPI, Celery, PostgreSQL, Elasticsearch, embeddings, hybrid retrieval, and human-review workflows. 📊 AI EVALUATION & RELIABILITY I care strongly about measuring AI quality instead of relying on “it looks good.” My work includes: - RAG retrieval evaluation - Context precision / recall / relevance - Groundedness and factual correctness - LLM-as-a-Judge - Agent trajectory and tool-call evaluation - Regression datasets - Langfuse tracing and observability - Human evaluation and failure analysis This is especially important for production agents, where the final answer can look correct even when retrieval, routing, memory, or tool usage failed internally. - What I Can Help You Build AI Agents & Automation · RAG / Enterprise Search · AI Chatbots & Copilots · Voice AI · Computer Vision · Industrial AI · Real-Time Video Analytics · Edge AI · AI Evaluation & Observability If you’re building an AI product - or already have one that needs to become more accurate, reliable, or production-ready - send me the problem. I can help take it from architecture and experimentation to a system people actually use.