Open to opportunities

Ramsha Anwar

@ramshaanwar

Message

Final-year Computer Science student building agentic AI and LLM systems.

Pakistan

Message

What I'm looking for

I’m looking to build agentic AI and LLM systems focused on reliable, observable behavior—multi-agent orchestration, grounded RAG, and rigorous evaluation. I want production deployment experience where measurement matters more than one-off demos.

I’m a final-year Computer Science student specializing in agentic AI systems, focused on building multi-agent LangGraph pipelines, RAG architectures, and fine-tuned LLMs. I enjoy going beyond demos by designing reliable, observable AI systems across healthcare and fintech domains.

I’m comfortable across the full stack for AI products—model fine-tuning and REST API design through production deployment and evaluation. I lean heavily into orchestration, evaluation, and observability, using tools like LangGraph, LangSmith, Ragas, and DeepEval to keep outputs grounded and measurable.

In my projects, I shipped an AI-powered SaaS with secure Supabase RLS and layered auth, built an agentic pharmacovigilance drug safety monitor with LangGraph + RAG grounded in cited sources, and developed an agentic trading pipeline with CrewAI and MCP-style tool decoupling. My goal is to keep iterating on agentic systems that are dependable in real-world conditions.

Experience

Work history, roles, and key accomplishments

Agentic Trading Intelligence System

Jan 2026 - Present (6 months)

Designed a multi-agent CrewAI trading pipeline with risk/critic agents that can veto flawed signals. Implemented a custom Model Context Protocol (MCP) server for market-data/broker tool access, added a backtesting framework, and deployed the system via Docker.

Python CrewAI Model Context Protocol (MCP)Pandas Multi Agent Systems Custom MCP Server Backtesting Frameworks Docker Risk Critic Veto Logic Koyeb Crew AI

AI Pharmacovigilance Drug Safety Monitor

Jan 2026 - Present (6 months)

Built a 4-agent LangGraph pipeline to flag medication risks using structured patient data and grounded retrieval from OpenFDA, DrugBank Open, and PubMed. Fine-tuned Llama 3 with QLoRA on FAERS data and instrumented groundedness/hallucination scoring with Ragas/DeepEval before deploying a FastAPI backend and dashboard.

Python LangGraph fastAPI ChromaDB GROQ LLaMA QLoRA OpenFDA DrugBank Open PubMed Sentence Grounded Evaluation Ragas DeepEval Koyeb Next.js

AI-Powered Job Seeker SaaS

OutSauce

Jan 2025 - Present (1 year 6 months)

Shipped a TypeScript/NestJS SaaS for job seekers and freelancers with Supabase-backed authentication, authorization guards, and Zod validation. Built RLS-enforced PostgreSQL data access, Stripe 4-tier billing with webhooks, and n8n automation for role-gated admin dashboards.