viditparashar.me

How I work

The first thing I add to an AI feature is a way to measure it. Without that, every prompt change is a guess delivered with a confident tone, and the team ends up arguing about vibes. Once there is an evaluation harness against real inputs, the arguments stop and the work gets faster.

The second thing is usually the escalation path. Deciding what the system does when it is unsure is a product decision disguised as an engineering one, and it is worth settling early — because a feature that abstains gracefully can ship, and one that guesses confidently cannot.

  • Measurement before opinions. A number ends an argument that a demo cannot.
  • Written updates over status meetings. If a week passes and I have told you nothing surprising, something is wrong.
  • I work in your repository and your tracker, not a separate place my work lives.
  • Narrow and working beats broad and demo-shaped. I would rather ship four actions that are reliable than forty that are shallow.
  • I will say when the honest answer is 'this is not worth what it costs'.

Related results

  1. viditparashar.me/faqviditparashar.me › faq › how-you-work

    How do you work day to day?

    FAQ

    Measurement before opinions. The first thing I add to an AI feature is a way to tell whether a change made it better, because otherwise every prompt edit is a guess…

  2. viditparashar.meviditparashar.me › experience › ainoviq

    Full-stack and AI Engineer at Ainoviq (Blackngreen) (Mar 2026 – present)

    Full-timeMar 2026 – present

    1 Mar 2026 — … and the automation that now handles hiring, sales outreach and LinkedIn. Also the person the company asks how to actually use AI in their own work. Built an AI…

  3. viditparashar.me/toolsviditparashar.me › tools › ai

    AI and LLM systems — tools Vidit actually reaches for

    10 tools

    AI and LLM systems: LLMs, LangChain, LLM orchestration, NLP pipelines, Model fine-tuning, Speech to text, Full-duplex voice, Agent orchestration, Workflow automation,…

Search this site for “How I work”

Vidit Parashar

Full-stack and AI engineer

I build AI features for domains where a confident wrong answer costs something — mostly clinical software and voice agents.

Gurugram, India · IST (UTC+5:30) · viditsharma818@gmail.com · @viditparashar96 · in/vidit-parashar-056707225 · viditparashar.framer.website

Summary

I have spent the last two years putting large language models into production rather than into demos. Most of that has been healthcare — an ambient scribe that turns a consultation into an EHR-ready note, and a set of clinical agents deployed into university health systems — where the interesting problem is never the model, it is what the system does when it is unsure.

Experience

Selected projects

Technologies

AI and LLM systems
LLMs, LangChain, LLM orchestration, NLP pipelines, Model fine-tuning, Speech to text, Full-duplex voice, Agent orchestration, Workflow automation, Screen understanding
Languages
TypeScript, JavaScript, Python, SQL, HTML, CSS, Java
Frontend
React, Next.js, Tailwind CSS, Redux, Redux Toolkit, Zustand, ShadCN, Material UI, NextUI, Styled Components, Bootstrap, GSAP, ApexCharts, Recharts, PWA
Backend and APIs
Node.js, Microservices, Socket.io, Azure Functions, Prisma, Mongoose
Databases
PostgreSQL, MongoDB, NoSQL
Infrastructure and tooling
Microsoft Azure, AWS, Docker, CI/CD, Git, Vercel, Firebase, Appwrite, Vite, Webpack
Mobile and desktop
Swift, SwiftUI, iOS (native), ElectronJS, React Native, Ionic

Education

Availability

open to interesting problems.