Vidit Parashar — Full-stack and AI engineer
Results
- viditparashar.meviditparashar.me › projects › conversational-agents
Conversational AI Platform — the layer other teams build voice and chat agents on
Case studyLive2026
1 Jul 2026 — Anyone building a conversational agent rebuilds the same substrate: speech in, an orchestration loop, tool calling, speech out, and some way to change a prompt without a deploy. Rebuilding it per product means each one gets a slightly different set of bugs, and the hard parts — latency, interruption, knowing when to stop — get solved badly several times over. A working platform for building voice and chat agents, with the turn-taking, tool-calling and configuration layers already solved rather than rebuilt per agent. It is running and you can try it — message me and I will give you access.
- Python
- TypeScript
- Next.js
- Node.js
- LangChain
- LLMs
- +4 more
- viditparashar.meviditparashar.me › projects › voxa
Voxa — dictation that learns your words, and an assistant that can see your screen
Case studyLive2026
30 Jun 2026 — Two things break the flow of working at a desk. Typing out something you could just say, badly and slowly — and having to describe what is on your screen to something that cannot see it. Dictation tools solve the first and then mangle every proper noun and piece of jargon you use. Assistants solve the second only if you paste in the context by hand, which costs more than it saves. Voxa runs as a working desktop app. Dictation learns the phrases and vocabulary you give it, “Hey Voxa” brings up an assistant that can read the screen and answer about it, and anything destructive or outbound goes through a confirmation step first. Wake-word accuracy sits above 95% in my own testing.
- 95%+wake-word accuracy
- 2personalisation surfaces
- 0irreversible actions without confirmation
- ElectronJS
- Node.js
- TypeScript
- Microsoft Azure
- Azure Functions
- LLMs
- +2 more
- viditparashar.meviditparashar.me › projects › ai-companion
AI Companion — a companion that listens and speaks at the same time
Case studyLive2026
1 Jun 2026 — Half-duplex voice assistants take turns: you speak, they wait, they answer. Real conversation does not work that way — people overlap, interrupt, murmur agreement while the other person is still talking. A system that can only take turns feels like a walkie-talkie, and every awkward pause reminds you what you are talking to. The companion holds an overlapping conversation rather than a turn-based one, and interruption behaves the way it does between two people.
- Python
- TypeScript
- Next.js
- Node.js
- LangChain
- LLMs
- +4 more
- viditparashar.meviditparashar.me › projects › medscribe
MedScribe — an ambient scribe that turns a consultation into a note an EHR will accept
Case studyLive2025
31 Dec 2025 — Clinicians spend a large part of every day writing down what just happened in the room. The tools meant to help mostly produce transcripts, and a transcript is not a clinical note — it has no structure, no separation between what the patient reported and what the clinician concluded, and nothing an electronic health record can file. So the documentation burden moves rather than disappears: someone still has to turn the wall of text into a SOAP note. MedScribe went into two large hospital chain clients across multi-specialty outpatient workflows, generating all three document types from live consultations.
- 2hospital chain clients deployed to
- 3clinical document types generated
- 2 monthsfrom empty repo to hospital deployment
- Next.js
- TypeScript
- Node.js
- Python
- LangChain
- LLMs
- +2 more
- viditparashar.meviditparashar.me › projects › clinical-agents
Clinical Agent Suite — four AI agents for the parts of care that are actually paperwork
Case studyShipped2025
31 Aug 2025 — A great deal of clinical time goes on work that is administrative rather than medical: writing the note, taking the intake history, looking up what the coding guidelines want, checking against guidance. Each of those is a different problem with a different tolerance for error, and treating them as one generic assistant produces something that is unreliable at all four. All four agents shipped and were integrated into EHR-compatible workflows, including in two academic health systems. The documentation pipeline cut physician documentation time by 40% in pilots — the number the clinicians themselves cared about.
- 40%physician documentation time saved
- 4clinical agents designed and shipped
- 2academic health systems integrated
- Next.js
- TypeScript
- Python
- LangChain
- LLMs
- Node.js
- +2 more
- viditparashar.meviditparashar.me › hire
Hire Vidit Parashar — open to interesting problems
ContactIST (UTC+5:30)
open to interesting problems. Email viditsharma818@gmail.com — replies within one working day, IST (UTC+5:30).
- viditparashar.meviditparashar.me › about › intro
Vidit Parashar — full-stack engineer, applied LLM and voice systems
I build AI features for domains where a confident wrong answer costs something. In practice that has meant clinical software — an ambient scribe, and agents for documentation, coding and decision support — and now a full-duplex AI companion and company-wide automation at Ainoviq, Blackngreen's AI division.
- viditparashar.meviditparashar.me › rates
Rates & engagement terms
Pricing
Rates are quoted per engagement after a short scoping call.
People also ask
There is no scrollback and no edit. Once something is spoken it is said, so the bar for what the agent will assert is higher — and the whole thing has to happen inside the pause a human leaves before assuming the line is dead.
Turn-taking is the part nobody expects to be hard. Deciding whether a pause means 'I have finished' or 'I am thinking' has no correct answer available at the moment you need it, and both mistakes are bad in opposite directions.
projects › conversational-agents
Conversational Agent PlatformVoxa and Resup.dev both run. Voxa I use daily — it is at v0.0.4; Resup.dev is live at resup.dev. Neither has an SLA behind it, and I say so on their pages rather than implying they are products.
They exist because a side project is the only place you choose the boring half yourself. Voxa's confirmation gate came from a genuine question — what do you do when the intent classifier might be wrong about sending an email — and the honest answer was to show the user the email.
projects › voxa
VoxaAlmost none of it is the model. It is the structure around it: knowing when the system should decline to answer, making a decision explainable after the fact, and choosing what happens at the edge of its competence.
The specific thing I have learned the hard way is that aggregate accuracy hides the failures that matter. A model is usually well calibrated on common cases and overconfident on rare ones — and rare cases are exactly where a clinician needed the help.
The fix is unglamorous: per-category thresholds, an explicit path to a human, and accepting a higher abstain rate rather than a better-looking headline number.
projects › clinical-agents
Clinical Agent SuiteMost of it, yes — that is a real narrowing and worth knowing. What it bought is that clinical software teaches you to design for being wrong, and that transfers to anything where a confident mistake has a cost.
The current work is voice agents rather than clinical, and my own projects are a desktop assistant and a résumé tool. But if you want someone with five years of e-commerce, that is not me.