User Journey

How ChatGPT discovers ucalyptus.me

A real conversation showing AI agent discovery of a personal site — from first query to deep profile understanding.

2:56
ChatGPT
SD
Step 1 · Discovery
https://ucalyptus.me/ search about this website

🔎 What the website is

ucalyptus.me is a personal portfolio belonging to Sayantan Das, a Machine Learning / AI engineer.

🧠 What the site contains

  • ML Research Engineer profile
  • Focus: LLMs Deep Learning PyTorch Transformers
  • Infra: Docker AWS Kubernetes
  • Links to GitHub, LinkedIn, projects

👉 Highlights

  • Worked at Prem Labs
  • Research at Vector Institute, ETH Zurich, ISRO
  • Built Text-to-SQL systems, Deepfake detection, LLM eval systems

⚙️ What it's for

Professional portfolio / resume hub — showcasing technical work, sharing projects, getting contacted for AI/ML work.

🌐 Sayantan Das 🤗 Hugging Face

If you want, I can break down his projects, strengths, or how strong this profile is for hiring.

Step 2 · Current Role
what does he do rn

👉 Senior Applied AI Scientist at Manulife

🧠 What that means

  • Works on AI / ML systems in industry
  • Site says "working to save lives" → tied to health / insurance analytics

⚙️ Also does alongside

  • Builds generative AI applications for businesses
  • Creates production-ready AI workflows and tools

📋 Before this

ML Engineer at Prem Labs — LLMs, reinforcement finetuning

🌐 Sayantan Das 🤗 Hugging Face

If you want, I can break down how strong his profile is compared to yours or hiring level-wise.

Step 3 · Research Deep-Dive
what are the papers he has worked on

📄 Papers by Sayantan Das

1. Kernel Space Conditional Distribution Alignment for Improving Group Fairness in Deepfake Detection
TMLR, 2025
Fairness in deepfake detection — bias mitigation across skin tones. Improves fairness vs accuracy tradeoff.
2. Unmasking Deepfakes: Masked Autoencoding Spatiotemporal Transformers for Enhanced Video Forgery Detection
IJCB, 2023
Vision transformers + spatial-temporal signals. Combines RGB frames + optical flow.
3. DARecNet-BS: Unsupervised Dual Attention Reconstruction Network for Hyperspectral Band Selection
IEEE GRSL, 2020
Remote sensing / satellite imagery. Selects spectral bands using dual-attention deep learning.
4. Topo Sampler: Topology-Constrained Noise Sampling for GANs
NeurIPS 2020 Workshop (Spotlight)
Generative models with topological constraints. Improves GAN sampling quality.

🧩 Research direction

Remote Sensing GANs → Deepfake Detection Fairness in AI

Moved from theory → high-impact applied research.

🌐 Sayantan Das 🌐 Sayantan Das +1

If you want, I can break down how strong these papers are tier-wise and what level he is at academically.

Message

This is an actual ChatGPT conversation discovering ucalyptus.me — showing how Agent-SEO makes a personal site fully discoverable by AI agents in 3 natural-language queries.