AI automation and marketing for businesses, from a Boston-based team working with clients across the US and beyond
We bring customers in through your website, search, ads and social media, then build AI systems that answer them, follow up and take care of routine operations.
DeepConnects is a Boston-based consultancy founded by Deep Shah and Darsh Chandura. We're a small senior team of AI engineers and automation consultants who have shipped production systems in healthcare, clean energy, education, media and operations. We test our methods inside a distribution business that Deep runs, so we build the way we would for ourselves and stay on to keep it running.
Deep Shah
Co-founder, AI systems and automation
Has shipped production AI across healthcare, media and operations: a WhatsApp intake system that cut a clinic's manual follow-up by 80%, an AI voice agent that calls every maternity patient at the right milestone, pediatric vaccine reminders running in multiple clinics, an OCR pipeline that has digitized 33 books, and a content platform that publishes itself. Also runs AI workflows and RAG tools inside a distribution business with 500+ retail accounts. Electrical engineer; previously co-founded an IT company.
Builds voice agents, RAG assistants and marketing automation, and measures each one against a baseline before it ships. At a US clean-energy company, his AI voice agents have handled 5,000+ calls and his weather-based ad model cut cost per click by 22% across more than $1M in ad spend. Before that he built AI assistants at Worcester Polytechnic Institute and a voice-based energy survey for Solomon Group, and he started out writing social content for national brands at FCB Kinnect. M.S. in Data Science with an AI concentration, WPI.
Marketing that brings customers in, and AI that handles the rest
Practical systems that are built, tested and handed over in a week or two, not months.
Marketing
AI-Ready Websites
New websites, and revamps of old ones, that people and AI assistants such as ChatGPT, Claude and Gemini can both read: real text, clean structure, structured data and fast pages.
Rank on Google, and get recommended when people ask ChatGPT, Perplexity or Google's AI Overviews. Technical fixes, content and tracking for both.
SEOGEOContent
Paid Ads
Google and Meta campaigns with budgets that follow real demand. A model Darsh built cut cost per click by 22% across more than $1M in ad spend.
Google AdsMeta Ads
Social Media
Posts planned, written and scheduled in your brand's voice, with AI handling the repetitive parts. Our content engine posts a Reel every day for @babyjunctions.
InstagramLinkedInFacebook
Video Production
Short-form social video, explainer and service videos, and product videos for your website, with AI doing the repetitive parts so you can publish often.
Automated pipelines that capture, qualify and follow up with leads on WhatsApp, so you never miss a hot prospect, even at 2 a.m.
WhatsApp APIn8nLead Scoring
WhatsApp Bots
Custom bots for appointment booking, customer inquiries, order updates and automated reminders. 24/7, with no staff required.
WhatsApp Business APIAI Replies
AI Voice Bots
Outbound and inbound AI calls in your patients' or customers' language: care reminders, follow-ups and confirmations, with a clean handoff to a person when needed.
Voice AITelephony
Private Internal AI Tools
Email briefs, daily reporting bots, dashboards and knowledge bases that run on your own infrastructure. Your data never trains a public model.
RAGPrivate AI
Process Automation
Connect your tools (ERP, spreadsheets, email) and automate repetitive workflows so your team stops doing the same steps by hand.
Make.comn8nERP Integration
AI Content & Document Systems
Self-publishing content engines and fault-tolerant OCR pipelines, proven on BabyJunctions and BookScan.
AI ContentOCR
Tools we work withGoogle AdsMeta AdsGoogle Analytics 4Search ConsoleSEMrushWordPressn8nMake.comWhatsApp Business APIPHPMySQLOpenAI APIGoogle SheetsNotion
Video Production
Video that shows what you do in seconds
People find and judge a business through short video now, on social feeds and on its own website. We plan, script and produce it with you, and use AI for the repetitive parts so you can publish every week instead of once a quarter.
What we make
Short-form social video. Reels, Shorts and TikTok clips with a hook in the first second, captions and your brand, made to a schedule.
Explainer and service videos. What you do, for whom and how it works, in a minute or two, for your homepage and service pages.
Product and demo videos. Your product or process shown clearly, for your site, ads and sales conversations.
Video engines that publish themselves. For steady output: scripts, visuals, music and posting automated, with you approving what goes out.
Proof you can check
A Reel every day, hands-off. Our content engine writes, renders and posts an animated Reel daily for @babyjunctions. See the project.
A YouTube channel that runs itself. For a client, AI creates the music, thumbnails and titles and uploads every video; 7 videos so far, 1,500+ views on the top ones. See the project.
Your team is doing work that software could do for them
Clinics, distributors, manufacturers and service businesses use DeepConnects to bring in customers through their website, search and ads. We also automate patient care, vaccine reminders, voice follow-ups, internal reporting and WhatsApp workflows, so they don't need to hire a tech team.
Hours lost copy-pasting between WhatsApp, Excel and your ERP.
Leads falling through the cracks
No system to track, qualify or follow up. Hot inquiries go cold.
Staff buried in repetitive tasks
Calling patients, sending reminders and chasing suppliers by hand, one at a time.
Why DeepConnects?
You have three options.
Freelance marketplace
Cheap and fast
Copy-paste templates, not custom logic
No understanding of your industry
One person, so one point of failure
Gone after delivery, leaving you to debug alone
Execution only, no strategy
DeepConnects
Senior, direct, fast
Built for your exact workflow
You talk to the people who build it
First version live in one to two weeks
Two founders and a senior team, so there is always cover
Marketing and automation in one place
A monthly engagement keeps it working and growing
Strategy and build, not just execution
Large agency
Expensive and slow
3 to 6 month timelines
Account managers between you and the builders
High overhead built into every invoice
Long contracts before anything is live
Overkill for most small-business automation needs
Industries
Industries we've built for
Clinics & HospitalsManufacturersDistributors & WholesalersService BusinessesClean EnergyHVAC & Home ServicesEducation
What Actually Gets Built
Systems that are running today
Zero missed inquiries
Lead Capture Bot
A WhatsApp flow captures, qualifies and notifies your team, even outside business hours.
Runs on its own
Auto Follow-Up Sequences
Multi-day WhatsApp sequences triggered automatically by where each lead is in the funnel.
No manual calls
Vaccine Reminder System
Tracks each child's vaccination schedule and sends parents WhatsApp reminders at the right time.
Fully automated
Patient Voice Bot
Outbound AI calls for prenatal guidance, postnatal follow-up and care pathway support, with no clinic staff effort.
No-shows drop
Appointment Bot
Patients book on WhatsApp, get confirmed automatically and receive reminders 24 hours and 1 hour before their slot.
Your stack, connected
Corporate AI Layer
ERP, CRM, accounting software and WhatsApp, unified into one AI layer with an internal knowledge base. Runs on your infrastructure.
33 books digitized
BookScan: Multilingual OCR
Turns scanned books in many languages and scripts into clean, searchable, structured text. A fault-tolerant digitization system that never loses a page, even across crashes.
39% more organic traffic
SEO & AI Search Content
Content planned and drafted with AI, edited by a person, published on a schedule and checked against what ranks on Google and what AI tools cite. The figure is from the content workflow Darsh built at WPI.
22% lower cost per click
Ad Budget Optimization
Models that move Google and Meta budget toward the days and places that convert. Darsh's weather-based version runs across more than $1M in clean-energy ad spend.
Publishes itself, daily
Auto Content & Instagram Engine
AI researches, writes and publishes articles, then creates and posts a daily Instagram Reel, with no manual work. Live: babyjunctions.in, @babyjunctions.
For Founders, CEOs & Business Owners
You run the business. AI handles the updates.
Private AI tools that summarize your inbox, report your numbers and put everything in one view, without sharing your data with anyone.
AI reads your inbox, filters what matters and sends a crisp daily summary. Your data never leaves your account.
No logins needed
Daily Reporting Bot
Sales, outstanding payments and inventory, delivered to WhatsApp every morning from your ERP or accounting software.
3 tools → 1 view
Central AI Dashboard
One private dashboard pulling live data from all your software. Ask questions in plain English. Runs on your server.
Instant answers
AI Knowledge Base
Upload your SOPs and past reports. Ask anything in plain English and get accurate answers from your own data.
Data Privacy, Non-Negotiable
Your business data stays yours. Always.
100% private
No AI training
Your data is never fed into any public AI model.
Your cloud
Your infrastructure
Systems run on your own server or existing cloud accounts, not a third-party platform.
Full visibility
No black boxes
You see exactly what data is accessed and how it is processed.
Who This Is For
This is built for you if…
You lose your first hour to emails and reports
You use three or more tools with no single view
You're uneasy about public AI platforms
You want AI on top of the systems you already have
You run a manufacturing, distribution or healthcare business
You don't have an in-house IT team
How It Works
Built for your exact setup. Live in under two weeks.
1
Tell us what you check every morning
2
We map your data sources and privacy setup
3
You receive the output. Nothing else changes.
Our Work
Built with AI. Tested in the real world.
Systems running in healthcare, distribution, manufacturing, content and consumer apps. Each one started with a job that had become too much to do by hand. The last six were built by Darsh in his roles at a US clean-energy company, Worcester Polytechnic Institute and Solomon Group.
Delivered, AI / Document Digitization
Surya OCRFastAPIReact + ViteCloud GPUWebSockets
BookScan: Multilingual OCR for Digitizing Scanned Books
A production-grade web app that turns scanned books in multiple languages and scripts, including older typefaces that defeat conventional OCR, into clean, searchable, structured digital text. Pages stream in live over WebSockets, shown side by side with the original scan for quick correction, and export as plain text, JSON or a searchable PDF with a real Unicode layer in each script's native fonts.
Live proof: 33 books digitized on the way to a 200-book library. Crash-resilient: a 500-page book interrupted at page 400 resumes from exactly where it stopped, even across server restarts.
Live editor: the original scan beside fully editable OCR text
The Problem
The goal was to digitize a large library of spiritual and philosophical literature, much of it printed in older typefaces, so it could be searched, studied and fed into AI systems. Conventional OCR breaks on these scripts and typefaces, and a multi-hundred-page book cannot afford to lose work halfway through.
What Was Built
BookScan pairs a modern vision-language OCR model (Surya) on a cloud GPU with a hardened processing pipeline. Users upload up to five PDFs and watch pages get recognized live, each shown beside its original scan with fully editable text. Every page is saved the instant it is recognized, so no work is lost; watchdog timers isolate problem pages so a single bad scan cannot stall a book; rendering was rebuilt around pypdfium2 to open each PDF only once, while batched GPU inference keeps the hardware fully used. The stack is deliberately pragmatic: a React + Vite frontend, a FastAPI backend on a rented GPU server, and SQLite for persistent per-book job history. REST and WebSockets let an inexpensive local machine drive serious GPU compute on demand.
Upload up to 5 PDFs (or process large books server-side) with a live per-book queue
Why It Matters
BookScan is the engine behind Sarathi, an AI-guided spiritual life-coaching app: it turns physical books into the high-quality, machine-readable corpus that powers retrieval and reasoning downstream. More than a wrapper around an OCR model, it is a complete, fault-tolerant document digitization system, designed for messy real-world inputs, multiple scripts and the unglamorous reliability needed to process thousands of pages without losing a single one.
AI Content GenerationSEO AutomationHands-OffPHPMySQL
BabyJunctions: Automated AI Content Platform
An AI content publishing system that runs on its own, with no manual work. Built as a live proof of concept at babyjunctions.in (199+ articles published, 2,500+ views). The same engine can be deployed for hospitals to publish health content automatically and rank better on Google.
Live proof: 199 articles published automatically, 763 products cataloged, 2,533 total views, with no content written by hand after the initial setup.
Sanatani Instrumental: AI-Automated Music YouTube Channel
Built for a client: a fully automated system that generates original spiritual instrumental music with AI, creates thumbnails and titles, and uploads videos to YouTube without anyone doing it by hand. The channel (@sanatani_instrumental) publishes regularly without ongoing work from the owner.
Live proof: 7 videos published, 1,500+ views on the top videos, all generated and uploaded automatically.
A mobile app for iOS and Android launching worldwide. Sarathi is an AI-powered spiritual guide that helps seekers explore dharma, spirituality and inner wisdom, available anytime, in the user's language, with grounded answers rooted in authentic tradition.
The AI engine simulator used to tune Sarathi: profiles, daily guidance, nightly trace and a feedback ledger
In Development, Pilgrimage App
iOSAndroidPilgrimage GuideJain TirthIn Development
Shatrunjay: Pilgrimage Guide Mobile App
A mobile app for the Shatrunjay Tirth, one of the holiest Jain pilgrimage sites. It pulls and structures content from shatrunjaydarshan.com to give pilgrims a detailed guide to the tirth (temples, routes, history and darshan timings), available offline on iOS and Android.
WhatsApp Business APIn8nPHPHealthcareLead Generation
Hospital WhatsApp Lead Generation AI
A multi-specialty clinic's reception team was losing patient inquiries during busy hours. Patients would message, wait hours for a reply and book elsewhere. We built a WhatsApp automation that captures every inbound inquiry instantly, qualifies patients, books appointments automatically and sends confirmation and reminder messages.
Result: 80% reduction in manual follow-up effort. Zero missed inquiries. No-shows dropped significantly with automated reminders.
The Problem
A multi-specialty clinic was receiving 30 to 50 WhatsApp inquiries a day. The reception team, also handling in-clinic patients, could not respond fast enough. Average reply time during peak hours: 2 to 4 hours. Patients who did not hear back simply messaged the next clinic. The business was losing 15 to 20 potential bookings every week to a slow inbox, with no way to track which inquiries had been followed up and which had gone cold.
What Was Built
A WhatsApp automation using the Business API and n8n that: (1) responds to every new inquiry in under 10 seconds, (2) walks the patient through qualifying questions: symptoms, urgency, preferred doctor and available time slots, (3) books the appointment and updates the clinic's system, (4) sends a confirmation, and (5) sends reminders 24h and 1h before the slot. Staff step in only when a query falls outside the automated flow.
Before
Receptionist handling 40+ WhatsApp threads a day by hand alongside in-clinic duties. 2 to 4 hour response times. 15 to 20 lost bookings a week. No follow-up tracking. High no-show rate.
After
Every inquiry answered in under 10 seconds. Appointments booked and confirmed automatically. Reminder sequences eliminate most no-shows. Reception staff focus entirely on in-clinic patients. 80% reduction in manual effort.
Key Lesson
The technology was the easy part. The harder work was mapping every edge case in the conversation flow: what happens when a patient says they are in pain right now, or asks something outside the bot's scope. A clean handoff to a person for those cases is what made the system usable in a clinical setting.
Hospital AI Voice Bot: Pregnancy & Post-Delivery Care
An AI voice bot that calls pregnant patients in their eighth month with personalized health guidance, and calls again after delivery for post-pregnancy care follow-up. No manual calls from hospital staff.
Result: 100% patient coverage on critical follow-up calls. Hospital staff freed from scheduling and making hundreds of individual calls each month.
The Problem
A maternity hospital wanted to give patients proactive care guidance: calling pregnant women at key milestones to tell them what to watch out for, and following up after delivery with post-pregnancy care instructions. In practice this was nearly impossible at scale. With hundreds of active patients, scheduling and making every call needed a dedicated staff member, and calls were often missed or delayed, which defeated the purpose of timely guidance.
What Was Built
An AI voice bot connected to the hospital's patient records. When a patient enters her eighth month of pregnancy, the system places a personalized voice call covering what symptoms to watch for, when to come in immediately and how to prepare for delivery. Once the delivery is recorded, a second call covers post-pregnancy care: breastfeeding guidance, recovery milestones, warning signs and when to bring the baby for the first check-up. Both calls use a natural-sounding AI voice in the patient's preferred language.
Before
Staff tracking patient milestones by hand and calling one by one. Calls often missed. Guidance varied depending on which staff member made the call. Hours spent daily on follow-up coordination.
After
Every patient gets her call automatically, at exactly the right milestone. Consistent, accurate guidance every time. No staff time spent on outreach calls. The hospital is seen as proactively caring for its patients.
WhatsApp Business APIAutomationPediatric Healthcare
Pediatric Vaccine Reminder System
An automated WhatsApp reminder system for pediatric hospitals. It sends parents the right vaccine reminder at the right time, based on the child's birth date and vaccination schedule. It runs in several pediatric practices.
Result: Vaccination compliance improved significantly. Parents receive timely, personalized reminders without any manual effort from clinic staff.
The Problem
Pediatric hospitals know that timely vaccination is critical, but reminding hundreds of parents individually, at the right point in the schedule, is hard to keep up with. Staff were checking records and calling or messaging parents by hand, which was slow, inconsistent and easily missed during busy clinic periods. Many parents missed vaccination windows simply because they forgot, not because they were negligent.
What Was Built
A WhatsApp automation that takes a child's birth date from the clinic's records and sends parents a reminder when each vaccine is due, based on the standard vaccination schedule. The reminder includes the vaccine name, what it protects against and a direct link to book the appointment. The system runs quietly in the background with no staff involvement once set up. Several pediatric practices now run it.
Before
Manual record tracking. Staff calling or messaging parents one by one. Reminders missed during busy periods. Parents relying on memory for vaccination schedules.
After
Reminders sent automatically at exactly the right time. Consistent messaging to every parent. No staff effort. Higher vaccination compliance. The hospital is seen as proactive and attentive.
A fully hands-off system that posts one animated slideshow Reel a day. AI writes the hook, on-slide text and caption; branded slides are rendered, music is added, and the Reel publishes to Instagram on a schedule, all on low-cost shared hosting.
Live proof: it runs @babyjunctions, posting a Reel a day with no manual effort.
Live, Lead Generation, built by Darsh at a US clean-energy company
PythonOpenAIMapboxNode.jsGoogle Business Profile API
HVAC Quote & Mapper: Instant Quotes for Homeowners
Homeowners enter their ZIP code and the system they need, and get an installation estimate in under 10 seconds, next to a map of the 500+ contractor service areas in the company's network. The tool also uses the Google Business Profile API to keep contractor locations and reviews in sync.
Result: 1,200+ homeowners quoted, a 28% lift in booking conversion, and the sales cycle cut from 14 days to 6.
Live, Paid Ads, built by Darsh at a US clean-energy company
Demand for heating, cooling and solar moves with the weather, so this model shifts Google Ads and Meta budgets based on the local forecast. It runs for five or more clean-energy clients.
Result: 22% lower cost per click and a 12% ROI lift across more than $1M in ad spend.
Live, Data Pipeline, built by Darsh at a US clean-energy company
PythonScrapySeleniumPostgreSQLAWS Lambda
AI Web Scraping Pipeline for a Product Marketplace
Collects product data from 50+ HVAC vendor websites, uses a supervised model to structure and categorize it, and copes with anti-bot measures and inconsistent page layouts. The output feeds a single consumer marketplace.
Result: 200,000+ products listed and 38% more marketplace transactions.
Delivered, Voice AI, built by Darsh at Solomon Group
PythonAlexaFine-tuning (DAFT)AWS LambdaDynamoDB
Mass Save Home Energy Survey, Rebuilt for Voice
Mass Save's home energy assessment survey, turned into a hands-free conversation on Alexa devices using domain-adaptive fine-tuning. Homeowners answer out loud and get personalized solar recommendations at the end.
Result: 300+ users in the first month, 34% more surveys completed, 18% more qualified leads and a 25% improvement in energy-optimization accuracy.
Delivered, RAG Assistant, built by Darsh at Worcester Polytechnic Institute
PythonLangChainLlamaIndexOpenAIAlexa Skills Kit
Campus AI Assistant with Alexa
A custom LLM with retrieval-augmented generation that answers student questions from the university's own information, as a campus chatbot and an Alexa skill. It was tested against an A/B baseline before rollout.
Result: 19% fewer help-desk requests.
Delivered, Student Wellbeing, built by Darsh at Worcester Polytechnic Institute
PythonLangChainSentiment ModelsFirebase
Mental Health Triage Agent
An AI agent for WPI's Center for Well-Being. It picks up emotional cues in what a student writes, offers guided self-help responses and connects the student with the counselor on duty when needed. Response quality was measured against a baseline, with agreed criteria for switching it off.
Result: 3,000+ conversations, and it contributed to a 27% drop in reported incidents.
Open for Work
Currently taking on new projects
The industries and kinds of projects we're taking on now.
Most practices focus on blocking or allowing AI crawlers with robots.txt rules, but that is the wrong problem to solve first. What matters is whether your website is structured so AI assistants can read it at all.
Meta's chief AI officer posted an AI-generated video of Sam Altman as part of a product promotion. It's allowed, it's labeled, and it signals something bigger: synthetic content is now mainstream marketing. For small businesses, this changes the equation.
Process
How we work together
From first call to live system
1
Free 30-minute call. Tell us the problem.
2
A custom roadmap: tools, logic, timeline and cost.
3
First version live in one to two weeks, then improved in real use.
After launch, there's a stabilization period when we monitor and fix issues at no extra charge. After that, a monthly engagement keeps the system maintained and growing, if you want one.
FAQ
Questions we get asked
We build AI-ready websites: new sites and revamps that ChatGPT, Claude and Gemini can read, understand and recommend. We also produce video (short-form social video, explainers and product videos) and run SEO, GEO, paid ads and social media. And we build AI automation, such as WhatsApp bots, AI voice agents, appointment reminders and lead follow-up, to handle what happens after a customer gets in touch. Our clients include clinics, contractors, distributors, manufacturers and service businesses.
Generative engine optimization: making sure your business shows up, and is described correctly, when people ask AI tools such as ChatGPT, Perplexity or Google's AI Overviews for a recommendation. It sits alongside regular SEO rather than replacing it.
Yes. We are based in Boston and work with clients across the United States, as well as the UK, Europe, Australia, New Zealand and the UAE. Most of the work happens remotely over email, video calls and WhatsApp, and our US line is +1 (774) 670-6255.
Book a free 30-minute call (Monday to Friday, 10 AM to 7 PM Eastern). Share what you want to automate and we will tell you the plan and the cost. No commitment.
A basic WhatsApp bot or lead generation workflow takes 1 to 2 weeks. An email brief for a founder can be running in under a week. Automation across several systems typically takes 3 to 6 weeks, including testing and handover.
Yes. Most modern systems have APIs or data export options that let AI tools connect. For older tools without APIs we use secure data extraction methods. Tell us your stack on the call and we will tell you up front whether it can be connected.
No. Every system is built with data privacy as the foundation. Your emails, reports and business data are processed privately, either on your own infrastructure or through secure APIs with no data retention. Nothing is fed to public AI training models.
Yes, that is the point. The systems we build are designed to stay out of your team's way. Your staff get a WhatsApp message, tap confirm or open a dashboard. They don't touch code, don't log into new software and don't change how they work.
Every project includes a stabilization period after launch when we monitor and fix any issues at no extra charge, and small adjustments in the first two weeks are included. For ongoing changes, new data sources and new automations we offer a monthly support retainer, so the system stays maintained and grows with your business. You are never left alone with something you do not understand.
The tools are the easy part. Most teams that try it themselves spend weeks learning the platform, build something fragile and abandon it when it breaks. What we bring is knowing which tools fit which problem, how to connect them reliably and how to design a workflow your team will actually follow. You get a working system in weeks, not a half-finished experiment.
For manufacturing, the most effective tools are n8n and Make.com for workflow automation, the WhatsApp Business API for customer communication, and custom dashboards for reporting. They integrate with existing ERPs and need no coding from your team.
Usually, yes. Tools like n8n are open source, so the running cost is mostly hosting. For a sense of what the right first project can do: the clinic lead system cut manual follow-up by 80% with zero missed inquiries, and the voice bot reached every patient without staff making the calls.
Let's Build Something
Ready to automate?
Tell us where you're losing leads or time. We'll tell you straight whether we can help, and what a first version would take. Free, no commitment.