class ShubhPathak:
def __init__(self):
self.role = "AI Engineer @ UBS Investment Banking"
self.location = "Pune, India"
self.focus = ["GenAI", "Multi-Agent Systems", "RAG", "LLMOps"]
self.building = "Production AI for regulated environments"
self.mantra = "Judgment over typing speed. Ship what runs."
def current_work(self):
return "A GenAI compliance platform processing 8,000+ docs/month"
def off_the_clock(self):
return "Tech Kahani - making ML accessible through story"- 🏦 Building a generative AI compliance-review platform at UBS that cut a 6-person team's manual review workload by 85%
- 🤖 I design agentic systems with planner / sub-agent orchestration, not chat demos
- 🧩 Comfortable across the whole stack: model layer, backend, frontend, and the database underneath
- 📖 On the side: Tech Kahani, a story-driven series that teaches ML to non-technical audiences
- 📫 Reach me at shubhpathakpro@gmail.com
|
A 9-agent LangGraph platform that automates the full ML workflow, from data profiling to deployment, with self-correcting execution and auto-retry on failure. Usable three ways: a frontend UI, a GitLab ticket-driven flow, and a pip-installable package.
|
An agentic workspace where an LLM plans and runs only the data tools a query needs, then returns a structured, sourced report. Tenant-scoped RAG on pgvector with database-tier row-level security.
|
|
My portfolio reimagined as a browser-based operating system: boot sequence, draggable windows, a live trainable neural network, a JD screener, and a working terminal.
|
A story-driven tutorial universe (the "Office Café") that teaches machine learning to non-technical audiences through recurring characters and narrative. A give-back project.
|
AI / ML / GenAI
Backend / Full-Stack
Data / Cloud / Infra
