Muhammad Ahmer S.

I build new AI products and get stalled prototypes into production: LLM agents, RAG, AI automation, AI integration, and computer vision. As an AI Engineer and Systems Architect, I own implementation, evaluation, and deployment. 11+ years in AI/ML engineering. 80+ production deployments. Expert-Vetted on Upwork, top 1% of talent. Claude Certified Architect, Foundations (CCAR-F). WHAT I CAN HELP YOU BUILD • AI agents and chatbots Large language model (LLM) application development, Retrieval-Augmented Generation (RAG), natural language processing (NLP), semantic search, tool calling, multi-agent orchestration, persistent conversation memory, and real-time voice. • AI integration and workflow automation Connect AI to your product, APIs, databases, and business systems, with validated inputs and clear permissions. • Computer vision and machine learning Deep learning for object detection, segmentation, OCR, multimodal image/video search, custom model training and fine-tuning, medical imaging, and edge AI. • Generative AI products Voice cloning, text-to-speech, AI avatars, image/video generation, and model-to-API delivery. • Production improvement and technical leadership Fix retrieval failures, inconsistent agent behavior, latency, rising inference costs, and deployment blockers. Architecture reviews and fractional AI leadership are also available. SELECTED WORK Biscuit AI | Retail shopping assistant Led five engineers on a shopping assistant deployed to in-store kiosks. Designed its RAG evaluation framework, session and persistent preference memory, and multi-agent architecture. My asynchronous execution refactor, integrated and tested by the team, brought representative response times down to about 2 seconds from 3 to 4 seconds. ShotDeck | Multimodal search and computer vision Built the AI layer for natural-language, visual, and color search, auto-tagging, and clip generation across 3M+ cinematic assets. A 20K similarity-curated training set outperformed a 100K random sample by over 20 percentage points on auto-tagging. Synthesys | Generative AI APIs Architected generative voice, face, and avatar APIs serving 2M+ calls annually at 99.9%+ uptime. Migrating to serverless GPU execution cut annual compute spend by approximately 80% from a $100K+ baseline. Shelfr | Retail computer vision Founding AI architect of a shelf-intelligence computer vision platform serving 200+ global brands. Built the perception platform from zero. Reported platform outcomes include +22% on-shelf availability and +25% shelf compliance. Shelfr received P&G’s External Business Partner Excellence Award. “Ahmer is very professional and managed to pull off quality work in a very short span of time. I would highly recommend working with him.” Karim Tribak, CEO & Co-Founder, Shelfr Edwards Lifesciences | Medical imaging research Developed echocardiogram de-identification for clinical research, masking patient-identifying regions while preserving diagnostic content before model training. HOW I WORK • Architecture The constraint chooses the architecture. I start with accuracy, latency, cost, privacy, and hardware requirements, then design the data pipeline, retrieval, models, integrations, and deployment around them. • Delivery and handover Evaluation, failure handling, observability, and machine learning operations (MLOps) are part of delivery. Your team receives documentation to operate the system. • Ongoing support Monitoring, retraining, and cost control as the system grows. • Scoping When scope is unclear, a short paid technical audit provides the blocker, recommended approach, and next steps in writing. SELECTED TOOLS AND PLATFORMS • AI and machine learning: Python, PyTorch, TensorFlow. • LLMs and agents: OpenAI API, Anthropic Claude, LangGraph, LangChain, LangSmith. • APIs and retrieval: FastAPI, Pinecone, PostgreSQL/pgvector. • Deployment: Docker, AWS, GCP, Modal, Replicate. I also design agent tool boundaries using Model Context Protocol (MCP). Message me or invite me to your job with the goal, current stack, and main blocker. I’ll tell you where I’d start and whether I’m the right fit.