ISHANT.OS — Agentic AI · Generative AI · Backend Engineering · Data Science

I BUILDSYSTEMSTHAT THINK.

Agentic AI, retrieval systems and the backend infrastructure that keeps them honest.

Interface initializing.

PROFILE GRAPH

Engineer working across agentic AI, generative AI, data systems and backend infrastructure.

Provisional bio — replace from the admin dashboard. Ishant builds systems where the intelligence layer and the infrastructure underneath it are designed together: agent orchestration, retrieval pipelines, and the databases, caches and APIs that make them dependable in production.

Agentic AI

Orchestrated multi-step agents with explicit state, tool boundaries and streamed reasoning.

LangGraphLangChainAgent OrchestrationTool Design

Generative AI

Retrieval-grounded generation, structured model output, and interfaces the model composes.

RAGEmbeddingsVector SearchStructured Output

Backend Engineering

Typed APIs, relational modelling, caching strategy and the operational edges around them.

PostgreSQLRedisGraphQLPrismaAPI Design

Data Science

Analysis and modelling in Python, with the pipeline work that makes results reproducible.

PythonNumPypandasMatplotlib

SKILL NETWORK

The edges matter more than the nodes.

Skills are stored with the relationships between them, so the graph shows what is actually used together — not a list of logos.

LANGGRAPHLANGCHAINAGENT ORCHESTRATIONTOOL DESIGNRAGEMBEDDINGSVECTOR SEARCHSTRUCTURED OUTPUTPOSTGRESQLREDISGRAPHQLPRISMAAPI DESIGNDEVOPSDOCKERMLOPSPYTHONNUMPYPANDASMATPLOTLIBTYPESCRIPTNEXT.JSAGENTICGENERATIVEBACKENDINFRADATAINTERFACE

Agentic AI

LangGraphLangChainAgent OrchestrationTool Design

Generative AI

RAGEmbeddingsVector SearchStructured Output

Backend Engineering

PostgreSQLRedisGraphQLPrismaAPI Design

Data Science

PythonNumPypandasMatplotlib

PROJECT MEMORY

Featured systems.

IN DEVELOPMENT

FOCAL

ISHANT.OS

A portfolio that an agent uses as its interface.

LangGraphRAGStructured OutputPostgreSQLTypeScriptNext.js

LIVE

QUEUED

VIS ERP

Production backend for an operations platform.

PostgreSQLRedisGraphQLPrismaAPI DesignDevOps

RESEARCH

QUEUED

Retrieval Engine

Grounded answers over a private corpus.

RAGEmbeddingsVector SearchPostgreSQLPython

INSPECTING SYSTEM

ISHANT.OS

A generative interface where the agent composes the page itself — validated UI specifications instead of generated code, streamed alongside a grounded answer.

InterfaceNext.js · GSAP · Motion · Lenis
Generative UITyped DSL · Zod validation · Component registry
AgentLangGraph · UI planner · Answer generator
KnowledgePostgreSQL · Vector index · Redis cache
  • Agent output is a validated specification, never executable code
  • Interface events are contract-typed and shared by mock and real backends
  • UI planning and answer generation run as separate stages

RESEARCH VAULT

Notes on making generated systems trustworthy.

RESEARCH

Generative interfaces without generated code

Provisional entry — replace from the admin dashboard. On treating model output as a validated presentation specification rather than executable code.

generative-uisafetystructured-output

RESEARCH

Grounding answers in a versioned knowledge base

Provisional entry — replace from the admin dashboard. On keeping retrieval, cache and index in step so answers never outlive their sources.

ragversioningcaching

AGENT RUNTIME

Query the system directly.

Commands resolve against portfolio content locally. `ask` hands off to the agent.

ishant.os — terminal
ISHANT.OS terminal — type `help` for commands

END OF SEQUENCE

Ask the system anything else.

OPEN TO SELECT WORK

Provisional — set your real availability in the admin dashboard.

KNOWLEDGE VERSION
seed-2026-09-12T14:59:53.717Z