I'm Ahmed Muharram, a senior at the University of Pennsylvania pursuing both Bachelor's and Master's degrees concurrently in Computer Science as an Onsi Sawiris Scholar. 🎓

• 🚀 I’m a software engineer who enjoys building end-to-end systems; especially where backend, data, and infra meet.
• 🤖 I’ve been working recently on LLM/RAG systems: everything from ingesting large sizes of data, hybrid retrieval, and grounding/eval, to making real deployments fast + reliable.
• 🧪 I’m also doing research on tool-calling LLM agents, focusing on how underspecified prompts and vague language affect planning, tool selection, and failure modes.
• 🐱 Fun fact: I’m a HUGE cat person. 🐾 I have 9 of them back home :)

• 🧠 AI/ML Engineer @ Fidelity Investments
  Scaled and productionized an internal LLM/RAG system: standardized connectors + ingestion templates, rebuilt retrieval for fast-updating corpora, and improved vector-store scalability.
• 📊 Data Science & AI Intern @ Fidelity Investments
  Built a high-throughput ingestion pipeline (TB-scale) and shipped grounded enterprise RAG workflows with evaluation + hallucination/grounding checks; deployed with Docker + Kubernetes.
• 🌾 Software Engineering Intern (AI Integration) @ Aydi
  Shipped an analytics chatbot for agricultural business owners (low-connectivity friendly), added testing + monitoring, and drove major LLM cost reductions through model/pipeline improvements.

• 🧾 Master’s Thesis: How Underspecified Prompts Shape Tool-Calling LLM Agents in Scientific Workflows
  Building agents + evaluation systems that turn open-ended questions into executable workflows, and measuring reliability/failure modes.
  • LAMINA: tool-augmented agent for scientific workflows (materials-focused)
  • PathFinder (in progress): conversational planning into step-by-step strategy graphs for pathogen data exploration, to be integrated later in VEuPathDB

• 🗣️ ScenIQ: Enterprise AI Meeting Intelligence (Senior Project)
  On-prem, security-first pipeline that turns meeting video into transcripts, topics, and structured action items (with a FastAPI + Postgres/pgvector backend and a React/TS frontend).
• 🔎 ByteNet: Distributed Web Search Engine
  Built a crawler + indexer + ranker on a custom distributed Java stack; implemented large-scale crawling, an inverted index, and PageRank + TF-IDF scoring with a query API.
• 🧠 PennOS: UNIX-like OS Simulator
  Implemented process abstractions, a preemptive scheduler, and a FAT16-style filesystem with a working shell.
• 📷 InstaLite: Instagram Clone w/ Semantic Search
  Full-stack social app deployed on AWS (EC2/RDS) with real-time chat + semantic profile discovery.

• 💻 Languages: Python, TypeScript/JavaScript, SQL, Java, C/C++
• 🧰 Backend/Infra: FastAPI, PostgreSQL/pgvector, Qdrant, Redis, MongoDB, Docker, Kubernetes, AWS (EC2, RDS)
• 🤖 LLM/Data: RAG pipelines, evaluation/grounding checks, Haystack, LlamaIndex, Spark; multimedia pipelines with FFmpeg + Whisper

• 🌐 Portfolio: https://muharram.dev
• 📧 Email: ahmed.o.muharram@gmail.com
• 💼 LinkedIn: https://www.linkedin.com/in/ahmed-muharram