Skip to content
Adarsh Jha
Adarsh Jha, AI Product Engineer

I build AI products that make it to production.

I work in the gap between application code and AI models: wiring LLMs into real product workflows, building evaluation pipelines, and shipping the UI that makes any of it matter. Currently at Humane Intelligence, where the red-teaming platform I work on is used by NIST, UNESCO, IMDA and the DoD.

View workRésumé
Currently

Software Engineer at Humane Intelligence — shipping red-teaming infrastructure used by NIST, IMDA, UNESCO, DoD

Selected work

Things I've built

AI security · 2026

MCP Scanner

Prompt-injection scanner for MCP servers, scored like a classifier

F1 0.800at 9.4% false-positive rate
TypeScriptNext.jsPostgreSQLValkeyBullMQGemini+1
AI / retrieval · 2026

Patently

Prior-art analysis that tells you which parts of your invention are already taught

2 LLM callsper analysis, any corpus size
TypeScriptNext.jsPython+5
Data engineering · 2026

MedSwitch

Finds a cheaper, chemically identical substitute for Indian prescriptions

₹5,684annual saving on one real prescription
TypeScriptPostgreSQLDrizzle+3
AI product · 2026

MetaGuard AI

AI data governance copilot for OpenMetadata

Next.jsTypeScriptOpenMetadata+2
Developer tools · 2026

DiffWatch

AI pull-request review with multi-persona feedback

<15sfrom PR link to full review
Next.jsTypeScriptGemini+2
All projects
Experience

Humane Intelligence

Sept 2024 — Present

Software EngineerCurrent

Build and ship red-teaming and evaluation infrastructure used by national standards bodies and government agencies.

  • Develop and deploy red-teaming projects across the platform
  • Integrate multiple AI model providers into the evaluation pipeline
  • Implement AI red-teaming features scoped to client requirements
  • Build the annotation workflow for validating model outputs and red-teaming results

Clients NIST · IMDA · UNESCO · DoD

Next.jsPostgreSQLDockerAWSVercel

1 Percent Verse

Feb 2024 — May 2024

SDE Intern

Built a React + TypeScript frontend for an AI-powered career roadmap platform (similar to roadmap.sh)

  • Translated Figma designs into production UI, rendering AI-generated learning paths and skill trees
  • Designed Supabase-backed REST APIs for user data, progress tracking, and roadmap delivery
  • Owned the integration layer wiring AI-generated content into the user-facing application
ReactNode.jsExpressTypeScriptSupabase
About

The interesting problems live in the gap between “the model returned something” and “the user got value.”

Most of what I do sits between application code and AI models — wiring LLMs into product workflows, building evaluation pipelines, designing multi-provider abstractions, and shipping the interface that makes any of it useful.

I care about shipping things that hold up: taking a fuzzy idea, picking a stack that fits the constraints, and getting a working version in front of someone fast enough to learn from it. Where a result can be measured, I'd rather measure it than argue about it — most of my recent projects ship with a labelled test set and a number attached.

AI / LLM

LangChainLangGraphLangSmithMCPRAGEvals

Languages

TypeScriptJavaScriptPython

Frontend

ReactNext.jsStreamlitTailwindMotionshadcn/ui

Backend

Node.jsExpressFastAPItRPCGraphQL

Database

PostgreSQLMongoDBRedis

Infra

DockerAWSVercelGitHub ActionsLinux
Live from GitHub

Recent activity

58 repos · @adarsh-jha-dev
Language mixTypeScript46%JavaScript35%Python7%HTML7%Jupyter Notebook2%
Contact

Let's build
something.

Open to full-time SDE roles from summer 2026, and available for contract work on AI products — LLM features, evals, agentic systems, developer tools. The fastest way to reach me is email.

adarshjhaxif@gmail.com
GitHubLinkedInX / Twitter