Projects

PilotKit

An insurance assistant that has to pass a test lab before it talks to customers. Jev decides and the LLM talks: every turn, Jev routes intent, catches prompt injection, flags PII and decides on human handoff before any tool runs, and gpt-4o-mini only extracts fields and writes the reply. Customers sign in to see their policies and file claims, shoppers compare 40 real plans, and voice input works too. In the lab, nine simulated customers (an angry claimant, a Hinglish speaker, a PII over-sharer, a prompt injector) chat with the bot and get graded against a rubric, and uncertain verdicts go to a human review queue. A turn costs $0.000359 with p95 latency of 4.23 s, and a full suite run costs about $0.02.

WebsiteSource

MKT·INTEL

An AI market research agent that sources every claim. A LangGraph team plans the research, searches and crawls the live web (Context.dev, then Tavily, then Firecrawl, with a credit breaker), extracts cited claims and has Jev make typed, confidence-scored decisions about each change, looping back when the evidence is weak. Eight playbooks, from pricing teardowns to TAM/SAM/SOM, export to PDF, PowerPoint, Excel, Markdown or JSON. Users can bring their own encrypted keys and models.

WebsiteSourceBlog

Jev Lab

Twenty-four decisions an AI product makes every day (classify, gate, route, filter, verify), each decided three ways: by Jev, by an LLM prompt and by plain code. Same labelled rows, real API calls, every mistake counted and charted as accuracy against cost per decision. Jev got 95.7% right across 486 decisions and lost 6 of the 24 projects. A full run of all 24 costs about $0.51.

WebsiteSourceBlog

AI-Powered Docker Repo Auditor

Six LLM agents read what three scanners found and grade a container image - CVEs, wasted layers, base image drift, CIS compliance. Four run concurrently under a per-agent timeout; two more consume the fan-in and see exactly which inputs can be trusted, so a failed agent degrades the report instead of quietly scoring you on less evidence. SQS FIFO worker on Fargate, live progress over Redis pub/sub, and a CI gate that blocks merges below 90% recall.

WebsiteSourceBlog

AZ-CLAW

A terminal-first AI coding agent that also runs from Telegram. Four modes - Agent, Plan, Ask and Second Brain - over one model factory that falls back from Groq to OpenRouter mid-call. Every file mutation is staged in memory and applied only after you approve a diff.

SourceBlog

GuardRAG

An enterprise-grade RAG chatbot with built-in compliance guardrails, multi-tenant architecture, and role-based access control. Supports PDF/DOCX/XLSX ingestion, namespace-isolated Pinecone vector storage, and a four-mode guardrails engine (Block, Alert, Approval, Redact) with full audit logging.

WebsiteSource

Hot Wheels x Pantone - Silver Series

An interactive, scroll-driven showcase for the Hot Wheels x Pantone Silver Series collection. Six die-cast vehicles, each paired with a distinct Pantone color, with deep-dive origin stories, full media galleries, and a Remotion-powered promo video.

WebsiteSource

LingoLearn

An AI-powered learning platform that transforms passive YouTube watching into active, quiz-driven education. It pauses videos at strategic moments to test comprehension, supports 130+ languages, and generates certificates on completion.

SourceBlog

Talent-Meet

Talent-Meet is a web platform that simplifies technical interviews with real-time coding, video calls, and structured candidate feedback in one unified experience.

Source

Vivah Styles

A staff-operated in-store platform for Indian wedding fashion boutiques - customer onboarding, AI-powered inventory and outfit matching, virtual try-on, billing, and sales analytics with role-based access for stylists, cashiers, and owners.

WebsiteSourceBlog