Kartikeya Parashar
Builder going deeper. Founder of 21 AI products at Kaydenlabs. Going from generalist to specialist in AI engineering.
Background
I've shipped 21 AI products over the last two years across fintech, legal tech, estate administration, content creation, social media, real estate, and tax. The flagships, WealthPilot (autonomous trading), Hermes Agent System (multi-agent OS), and LastingPath (estate admin, live at lastingpath.com), are where the depth lives. Currently VP Applied AI at Onclave.
Last month I realized I was a mile wide and an inch deep. I know how to ship an LLM product. I don't yet know how to architect one for a Fortune 500 customer with security, compliance, and SOC2 constraints. So I'm spending the next 12 months going deeper. Field Notes is the public layer.
Why Forward Deployed Engineering
Shipping 21 products taught me how systems break in production. The retrieval pipeline that worked in dev, the agent loop that drifted under real customer data, the kill-switch you needed at 2am. I want to do that work for customers operating at scale instead of for my own portfolio. Forward Deployed Engineer is the role that maps to that ambition: sit with the customer, find what is actually broken, build the system that fixes it, take the production pager.
Currently learning
- Evals discipline (currently building a 50-prompt eval suite for one of my products)
- Production RAG patterns (hybrid retrieval + reranking + retrieval/generation eval separation)
- Anthropic engineering practices (cookbook, Building Effective Agents, Contextual Retrieval)
- LLM internals (Karpathy, Jay Alammar, Toy Models of Superposition)
- Classical system design at intuition level (~6 canonical patterns)
Flagships
Full portfolio →Elsewhere
See the full portfolio: Kaydenlabs / Work →
Current role: VP Applied AI at Onclave.
The work is the goal. The doors open from the work.