Document & Extraction Pipelines
Multi-model document extraction pipelines that read long, dense documents, pull out the structure, and cite every line back to its source.
I build AI into the places where a wrong answer has real consequences. The systems live in regulated environments, designed to flag what they are unsure about and route the rest to a human.
I am Kalpesh. I design, build and ship production AI for organizations with complex or regulated workflows. One person, from the architecture to the running system, with nothing handed to a bench you never meet.
A demo calls an API. Production is everything around it: the pipeline, the guardrails, the human handoff, and the system it lives inside.
Multi-model document extraction pipelines that read long, dense documents, pull out the structure, and cite every line back to its source.
Rule plus LLM hybrids that check submissions against your guidelines and route only flagged cases to a human.
Scoring and recommendation engines that show their reasoning, so they support the decision instead of hiding it.
Salesforce AI development and LLM features surfaced inside the tools your team already uses, so there is nothing new to learn.
Built for regulated and high-stakes work, six live in production and two voice systems in development. Client names withheld. Same operator, different shapes of problem.
A two-model pipeline reads government RFP PDFs and extracts requirements, deadlines, and rubrics. A second model reviews and flags the extraction.
A rule plus LLM hybrid validates submissions against long guidelines and marks each requirement pass, fail, or needs a human.
Scores cases against a structured catalog of evidence-based therapeutic models, recommends the best fit, and shows its reasoning. The decision stays human-led.
A multilingual voice front-end on a 311 line, all-AWS over a Salesforce routing brain. Routine requests become cases and are deflected; unclear or urgent calls transfer to a live agent — one engine shared with the on-screen console.
An all-AWS generative voice AI on an emergency intake line. Every call is treated as an emergency by default; confident non-emergencies are warm-handed to the right service, and the dispatcher sees read-only unit and ETA data. The AI never allocates a unit.
Beyond client work. Products designed, built, and run end to end, on my own.
A zero-knowledge password + passkey manager for teams. Save, autofill, and generate logins, store passkeys, and share any login with a team — end-to-end encrypted, with roles and real revocation.
A Chrome extension that scrapes business leads straight from Google Maps — name, phone, website, category, reviews — and exports clean CSVs ready for outreach. Free to 40 leads, Pro unlimited.
Phone-triggered remote ignition for Claude Code sessions. Tap from your phone, an agent on the home machine boots a fresh session from zero and hands back a live link. Pull-based, so the cloud never reaches into the house.
Plan a posting rhythm, queue shorts and long cuts together, and publish to YouTube with thumbnails and playlists in one pass. The video never touches my servers: it uploads from the browser straight to YouTube.
A live team Pomodoro product. Shared focus rooms, ambient sound tools, a full marketing site, and three free tools. Designed, built, and maintained solo.
Start with the problem, not the tech. What is blocked, what a wrong answer costs, and where a human stays in the loop.
Design the pipeline, guardrails, and interface, then build end to end. Prompts, infrastructure, and the system it lives inside.
Ship into the real environment, verify the behaviour, and hand over a system your team can run without me, not a demo that needs babysitting.
A 15-minute call about the problem you are trying to solve. We work out together whether AI is the right tool, and what shipping it would actually take. If it is not, I will say so.