AI engineering & agents
Building AI systems that work beyond the demo — agentic workflows, retrieval-augmented generation, tool-use patterns. The hard part isn't calling the API, it's making it reliable.
A working invitation
I take on a few engagements at a time — where I can own a real piece of the problem and ship something that has to work on a Tuesday morning, not just in a demo.
What I do
Building AI systems that work beyond the demo — agentic workflows, retrieval-augmented generation, tool-use patterns. The hard part isn't calling the API, it's making it reliable.
Event-driven architectures, database design, queue-based workflows. Systems that handle failure gracefully and scale without drama.
From schema to screen. TypeScript, React, Node.js, PostgreSQL — shipping end-to-end, not just one layer.
The decisions you make in week one compound for years. I help teams get the foundations right — or fix the ones that aren't.
Where I’ve shipped
Five years of production code across four industries. Each taught me something the next couldn’t.
Cerner (Oracle Health), Clipboard Health
FHIR compliance, clinical data pipelines, EHR integrations.
CommandK
GDPR, DPDP, PII detection and automated remediation.
Mensari
Transaction indexing, DAO governance, on-chain analytics.
Saltmine
Space planning, capital projects, complex domain modelling.
The path
Tell me what you're building and what's broken.
You describe the problem, I ask uncomfortable questions.
Scope, approach, timeline, rate. No surprises.
Send a message
Fields marked with asterisk are required.
Your email client should have opened with the message. If it didn't, you can reach me directly at hello@shaileshshivam.dev.
Before you ask
Agentic systems, RAG pipelines, LLM integration into real products. I focus on the gap between 'works in a notebook' and 'runs in production' — structured outputs, eval frameworks, cost management, graceful degradation when models fail. I use the Anthropic and OpenAI APIs, LangChain when it fits, and raw API calls when it doesn't.
TypeScript and Node.js for most backend/full-stack work. Python for ML and data pipelines. PostgreSQL, Redis, AWS. React on the frontend. I don't chase frameworks — I pick what's proven for the problem and what the team can maintain after I leave.
IST (UTC+5:30), based in India. I've shipped production code async with teams across US, Europe, and Asia — timezone hasn't been a bottleneck.
2+ weeks for project work. Shorter for architecture reviews or advisory. I scope tightly and deliver in iterations, not big reveals.
The case studies at /work show real systems — a PII detection engine using LLMs, event-driven blockchain indexers, zero-downtime migrations at scale. Happy to walk through specifics on a call.