Islamabad, Pakistan · Remote since 2019

Saad Arshad

I develop production AI systems that read, reason, and act — powered by LLMs, agents, and retrieval.

Scroll
01 · About

NLP, agentic AI, and production-scale systems.

Highly accomplished AI/ML Engineer with over 6 years of experience specializing in Natural Language Processing, Agentic AI, and Large Language Models. I focus on the end-to-end design, development, and deployment of scalable, production-ready AI solutions.

I'm currently AI (NLP) Technical Lead at Intelligent Relations, where I lead the architecture of Preston, an AI-powered PR automation platform. I thrive in async-first teams and have worked remotely with international companies since 2019.

6+
Years in NLP / LLMs
9–10M
Articles processed / month
500K+
Journalist profiles
900+
Category taxonomy
02 · Experience

Career timeline.

MAR 2024 — PRESENT

Intelligent Relations

AI (NLP) Technical Lead · Remote

  • Led the end-to-end architecture and deployment of Preston, an AI-powered PR automation platform delivering multi-agent chat, automated journalist outreach and replies, media monitoring, semantic search-based recommendations, and personalized pitch generation.
  • Built a multi-agent chat interface enabling users to perform any platform action through natural language — searching journalists, launching campaigns, monitoring coverage, and generating pitches.
  • Designed a campaign creation workflow that reduced manual effort by 90%, automating journalist targeting, pitch drafting, and follow-up strategy.
  • Built PR Deep Dive, an agent-based research automation system that cut manual research effort from hours to minutes.
  • Developed a hierarchical categorization system spanning 900+ categories, improving search relevance by 25%.
  • Managed the full MLOps lifecycle across AWS (Fargate, Lambda, EC2), Docker, and FastAPI.
JUN 2022 — MAR 2024

Intelligent Relations

ML/NLP Engineer · Remote

  • Engineered high-throughput NLP pipelines analyzing 9–10 million articles per month to build dynamic profiles for 500,000+ journalists, publications, and podcasts.
  • Developed an AI pitch generator that analyzes successful pitches to generate on-brand outreach emails, reducing pitch writing time by 80%.
  • Built a journalist recommendation engine that dynamically updates query representations using user feedback and engagement signals.
  • Developed Chat with Preston, an LLM-powered assistant for complex PR-related queries.
NOV 2021 — JUN 2022

Lahore University of Management Sciences (LUMS)

Graduate Research Assistant · Onsite

  • Collected and curated the first benchmark dataset for Urdu abstractive news summarization — 180,000 news and summary pairs.
  • Performed lossless compression and language adaptation of the mT5 model, reducing parameters by 50% while maintaining accuracy.
  • Pre-trained mT5 with a denoising objective and fine-tuned it for abstractive summarization, significantly improving ROUGE scores.
SEP 2020 — NOV 2021

Vertical Solutions

ML/NLP Engineer · Remote

  • Trained multi-label text classification models to categorize legal invoice clauses.
  • Developed a scoring and recommendation pipeline for legal contract clauses.
  • Fine-tuned semantic similarity models to detect and correct wrong legal codes.
MAY 2019 — SEP 2020

Automotive Artificial Intelligence (AAI) GmbH

ML/NLP Engineer · Remote

  • Developed and deployed a domain-specific conversational AI system with integrated speech recognition and synthesis.
  • Managed API development and deployment on AWS infrastructure.
  • Fine-tuned object detection models for industry-specific automotive use cases.
03 · Education

Academic background.

Lahore University of Management Sciences (LUMS)

2020 — 2022

MS in Computer Science

Thesis: "Urdu Abstractive News Summarization" — curated a 180k-pair dataset, pre-trained & compressed mT5 (50% parameter reduction) with preserved ROUGE scores. GPA: 3.84 · Major: AI, NLP · Dean's Honour List.

National University of Computer & Emerging Sciences (FAST)

2015 — 2019

BS in Computer Science

GPA: 3.59 · Dean's List: Fall 2015, 2017 & 2018 — Spring 2018 & 2019.

04 · Projects

Selected work.

PR Agent

An autonomous agent that runs the full PR cycle end to end — media monitoring, journalist targeting, pitch drafting, and outreach — producing weekly briefs and campaign-ready recommendations with minimal human input.

PReston

AI-powered PR automation platform delivering multi-agent chat, automated journalist outreach, media monitoring, semantic search-based recommendations, and personalized pitch generation — cutting campaign workflow effort by 90%.

KeyCrew

Led development of KeyCrew's AI article generation system, producing perspective, feature, consumer, and press release content from interview transcripts, plus a sales intelligence pipeline identifying prospects from article content and expert mentions.

Visit KeyCrew

Authr

A content-intelligence SaaS that crawls a brand's blog and competitor blogs, clusters content into a vector space, identifies coverage gaps, clones brand voice, and drafts SEO-linked posts.

Visit Authr

Fake News Detection (USE)

Fake news detection using Google's Universal Sentence Encoder and a simple neural network.

View on GitHub

Chatbot (Seq2Seq)

A chatbot developed using a seq2seq encoder-decoder architecture.

View on GitHub
05 · Skills

Tools & technologies.

AI / NLP / LLMs
Large Language Models Transformers RAG Semantic Search Embeddings AI Agents Prompt Engineering LoRA / PEFT Model Quantization
MLOps & Serving
Docker FastAPI Model Deployment Monitoring & Observability
Frameworks & Libraries
PyTorch Hugging Face LangChain Scikit-Learn SpaCy Pandas
Data & Infrastructure
AWS EC2 / S3 Bedrock Fargate / Lambda PostgreSQL MongoDB Pinecone ChromaDB
06 · Publications

IEEE research.

CodeBERT for Code Clone Detection: A Replication Study
Saad Arshad, Shamsa Abid, Shafay Shamail
IEEE 16th International Workshop on Software Clones (IWSC), 2022
A Comparative Analysis of Clone Detection Techniques on SemanticCloneBench
Sohaib Masood Rabbani, Nabeel Ahmad Gulzar, Saad Arshad, Shamsa Abid, Shafay Shamail
IEEE 16th International Workshop on Software Clones (IWSC), 2022
07 · Contact

Let's build something.

Open to conversations about AI engineering roles, collaborations, or anything agentic-AI related. Reach out directly — I read every message.

saad.arshad102@gmail.com