Aaditey Pillai

About Aaditey Pillai

I’m an engineer who loves turning messy, real-world problems into clear, working systems. I started in electronics and communication at SRM, where I learned to think in signals, constraints, and trade-offs. Somewhere between late-night labs and scrappy side projects, I realized what really draws me in is building intelligent tools that make life simpler, especially when the data is imperfect and the requirements move. My first real spark for AI came during a cybersecurity internship. I was responsible for network surveillance which meant watching patterns, and responding fast. That’s where I met Palo Alto Networks’ XSOAR. Automating playbooks and seeing “smart” response loops in action gave me a first taste of what AI-driven systems could do. That experience flipped a switch for me, and I haven’t looked back since.🤖 At Duke, I’ve been exploring that curiosity at scale - learning how large language models, retrieval, and agents actually behave once you wire them into a product. I like the whole lifecycle: framing the problem, shaping prompts and schemas, testing failure cases, instrumenting evaluation, and iterating until the thing is so reliable people forget there’s “AI” inside and just get their work done.🧪 Across internships, I picked up range and discipline. Cybersecurity taught me to respect uptime, logs, and alert fatigue which are the quiet, unglamorous things that make systems trustworthy. In data-heavy work, I learned to love clean pipelines, careful feature work, and improvements that actually move the metric that matters. More recently, I’ve been hands-on with LLMs in production - designing prompt architectures, extending extraction pipelines, and building evaluation loops so we can prove progress, not just claim it.🧰 What I optimize for: useful > flashy. Models cheap enough to run, prompts simple enough to maintain. I like pairing with product and domain experts, writing crisp docs, and leaving teams stronger than I found them. If you’re working on trustworthy AI systems, document intelligence, agentic workflows, or you just want to jam on evaluation and prompt design, I’d love to connect. Let’s build tools that are quietly excellent and obviously helpful.

Experience

1. Ai Engineer

   PRGX Global Inc.

   Aug 2025 — Present · Atlanta, GA, US

   Evaluate open-source LLMs for contract intelligence: Benchmark GPT OSS against current models for entity extraction, prompt engineering, quantization, and few-shot prompting to optimize accuracy and cost-efficiency.2. Develop and test scalable AI/LLM pipelines: Implement text matching, fuzzy search, and contract grouping rules, while validating performance with unit tests and JSON-based datasets.3. Design and optimize NLP-driven contract analytics: Build and refine matching logic and hierarchy models (e.g, linking MSAs, amendments, and SOWs) to improve contract risk scoring and document relationships.

Education

• Duke University

  Master of Engineering - MEng

• SRM IST Chennai

  Bachelor of Technology - BTech