Maryam F.

Top Rated Plus - 7+ Years on Upwork I build AI systems that actually answer correctly, not chatbots that hallucinate and embarrass you in front of users. RAG pipelines, multi-agent workflows, and LLM apps that ground every response in your real data. If your AI needs to be reliable enough to put in front of customers or run inside your business, you're in the right place. Most "AI projects" fail for the same reason: a demo that works once gets shipped as a product, then falls apart on real queries. I build for the second case, not the first. What I do: - RAG systems: retrieval pipelines that pull the right context every time, with chunking, re-ranking, and citation so answers are grounded and traceable - Multi-agent orchestration: LangGraph-based agents that plan, call tools, and hand off work, built to recover from failure instead of looping forever - LLM integration: OpenAI, Anthropic (Claude), and Hugging Face models wired into your stack with cost and latency controls - Vector databases: Pinecone, Weaviate, Qdrant, Chroma (schema design, embedding strategy, hybrid search, index tuning) - Prompt engineering: structured prompts, evals, and guardrails that make output consistent instead of lucky - Fine-tuning: domain adaptation when prompting isn't enough, with honest advice on when it's worth the cost and when it isn't - NLP: spaCy pipelines for extraction, classification, entity recognition, and text processing at scale - Embeddings: model selection, chunking strategy, and retrieval tuning for accuracy that holds up in production My stack at a glance: Python, LangGraph, LlamaIndex, OpenAI, Anthropic, Hugging Face, Pinecone, Weaviate, Qdrant, Chroma, spaCy, embeddings, prompt engineering, fine-tuning, multi-agent orchestration, RAG, NLP How we'll work: - We talk through the real problem (not just "add AI to it") - I scope it honestly: approach, timeline, cost, and where the accuracy risks are - I build in clear milestones with evals so you can see it actually working, not just demoing - You get clean, documented, tested code you own completely Ready when you are.