Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up
Appearance settings
@ishita1989
ishita1989
Follow

Ishita Shah ishita1989

Exited Founder & AI Product Leader Building Safety First Agentic Systems in Production
  • SF Bay Area

Block or report ishita1989

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ishita1989 /README.md

Hi, I'm Ishita 👋

AI Product Leader · Exited Founder · Production LLM Systems · Evals, Safety & Agentic Workflows

I build AI products that have to work in the real world — not just in demos.

My work sits at the intersection of product strategy, applied AI, user trust, and production reliability. I care about building AI systems that are grounded, evaluated, observable, and safe enough for users to rely on when the stakes are high.

Most recently, I founded and led Spellbound, an LLM-powered fertility and pregnancy guidance platform that scaled to 100K MAU, became cash-flow positive within 12 months, was adopted by 10+ clinical partners, and exited to a healthcare chain.

I use this GitHub to share sanitized product case studies, AI workflow patterns, and prototype systems that reflect how I think about building reliable AI products.


How I build AI products

I like building from first principles:

  • Start with the user workflow and the real failure modes
  • Define what the product is — and what it should never try to be
  • Treat prompts, retrieval, memory, and evals as product surfaces
  • Design evals and safety gates before scaling
  • Use human-in-the-loop workflows where the AI should not act alone
  • Measure quality in production, not just in demos
  • Iterate across UX, prompts, retrieval, routing, safety, and model behavior

I’m strongest at the product-engineering boundary: translating ambiguous user problems into AI workflows, making technical trade-offs explicit, and partnering deeply with engineering to ship systems users can trust.


What I'm working on now

I’m currently building and documenting production-inspired AI prototypes that explore how to move from impressive demos to reliable, high-trust AI products.

  • 🧪 AI evals before launch — designing eval harnesses, synthetic user scenarios, LLM-as-judge rubrics, and release gates for agentic workflows
  • 🛡️ Safer RAG systems — exploring retrieval quality, source grounding, hallucination checks, confidence gating, and citation-aware responses
  • 🤝 Human-in-the-loop AI — designing escalation paths for moments where the AI should clarify, defer, or route to a human instead of acting alone
  • 🧭 Agentic workflow design — turning ambiguous user goals into structured AI workflows with clear boundaries, tool use, memory, and fallback behavior
  • 📊 Production AI observability — thinking through telemetry, feedback loops, drift signals, rollback triggers, and quality monitoring after launch
  • ✍️ Writing about applied AI product work — sharing notes on what breaks, what needs to be measured, and how product teams can build trust into AI systems from day one

Selected work

Spellbound

Commercial product · Not open-source

An LLM-powered fertility and pregnancy guidance platform for high-trust, medically sensitive user journeys.

Product/system areas:
RAG · Clinician-verified knowledge base · OCR report interpretation · Model routing · LLM-as-judge evals · Safety guardrails · Human escalation · AI observability

Impact:
Scaled to 100K MAU, adopted by 10+ clinical partners, became cash-flow positive within 12 months, and exited to a healthcare chain.


BlitzPrep

Commercial product · Not open-source

An AI job-search and interview-prep copilot designed to help candidates move from generic applications to targeted, high-conversion job-search execution.

Product/system areas:
Semantic job matching · Resume tailoring · Interview simulations · Hiring-manager research · Role-specific rubrics · Candidate feedback loops · Personalized prep workflows


📊 [Brand OS] — Strengthen your brand on LinkedIn by creating and posting consistent content in your unique voice as well as improve it by tracking its performance

🛡️ [Trade X] — Deep equity research and automated trading agent

🧪 [RocketShop] — Create brand content for your D2C store

🎯 [Sureshot AI] - Initial version of BlitzPrep


Focus areas

  • AI systems: Agentic workflows, RAG, multimodal AI, model routing, personalization
  • Quality & safety: LLM evals, LLM-as-judge, hallucination checks, guardrails, red-teaming, human-in-the-loop review
  • Production AI: Observability, telemetry, feedback loops, rollback triggers, cost/latency tradeoffs
  • Product: 0→1 strategy, PMF discovery, product roadmap, GTM, experimentation, cross-functional execution
  • Technical: SQL, Python, prototyping, prompt/workflow design, AI system debugging

About me

  • Exited Founder & Head of Product at Spellbound (acquired Feb, 2026)
  • Previously Senior AI PM at Magic Technologies and AI PM at KPMG India
    • Early career Software Engineer at Juniper Networks
  • Carnegie Mellon University, MS in Software Management
  • Based in San Francisco Bay Area

Let's connect

💼 LinkedIn
📧 ishita@ishitashah.me

Popular repositories Loading

  1. branding_os branding_os Public
  2. Hands-On-Large-Language-Models Hands-On-Large-Language-Models Public

    Forked from HandsOnLLM/Hands-On-Large-Language-Models

    Official code repo for the O'Reilly Book - "Hands-On Large Language Models"

    Jupyter Notebook

  3. TradeX TradeX Public

    TradeX — Deep Equity Research & Automated Trading Agent. Full-stack platform with AI-powered research (DCF, moat analysis, thesis generation) and multi-strategy trading engine (momentum, event-driv...

    Python

  4. academia_brain academia_brain Public

    Collaborative Research Brain — clickable Next.js prototype of an AI-native research intelligence platform for academic labs. Five UX surfaces + context-aware chat drawer. Railway-ready.

    TypeScript

  5. open-design open-design Public

    Forked from nexu-io/open-design

    🎨 Local-first, open-source alternative to Anthropic's Claude Design. ⚡ 19 Skills · ✨ 71 brand-grade Design Systems 🖼 Generate web · desktop · mobile prototypes · slides · images · videos · HyperFra...

    TypeScript

  6. ishita1989 ishita1989 Public

    Profile README

AltStyle によって変換されたページ (->オリジナル) /