Can Bharosa’s Voice-Based AI Be The Healthcare Gamechanger For Tier II & III India

The pain was intense, often excruciating, during her pregnancy. Miti Jain was laid up because the recurrent pain in the stomach and in her back could not be diagnosed and healed even after visiting several doctors. 

His wife’s plight through that crisis, uncertainty, and stalemate brought Vandit Jain in front of a glaring disconnect in India’s booming healthcare sector. It was the lack of clarity in communication. “It’s difficult, if not impossible, for patients who are not equipped with enough knowledge in medical science to comprehend what was wrong with them,” he said.

“What makes the situation grimmer is the diversity in our languages. There’s often a huge gap between understanding what exactly a patient is suffering from and what exactly the doctor is advising. This eventually leads to a loss of faith in the system and greater suffering for the patient.”

The seeds of Bharosa, the Hindi word for faith, had germinated when Jain shared his thoughts with a friend from the medical fraternity, Dr Amit Gupta. 

In his years of clinical experience, Dr Gupta had seen firsthand that for countless patients across the country, getting timely, clear answers, especially while seeking a second or a third opinion, remained largely out of bounds. Millions are left in the lurch without proper guidance and, in fact, most people do not have access to proper doctors. The faith in healthcare was on the wane, lamented the founder. 

The rise of artificial intelligence (AI) in India’s $638 Bn healthcare sector complemented a surge in investment and addressed the severe dearth of medical professionals to a large extent, besides speeding up the treatment, providing smarter assistance in various clinical procedures, and taking healthcare facilities to underserved rural India. 

From advanced AI conversational platforms to AI-based symptom checkers and multilingual voice-based health assistants, startups have swarmed in to solve critical pain points along the journey, paving the way for a $758 Mn AI-assisted healthcare market to reach $8,728 Mn by 2030. Players such as mfine, HealthifyMe, Tata 1mg, Practo, Jivi.ai, and Dozee are leveraging AI in different ways to improve diagnosis, patient engagement, chronic care management, and hospital monitoring. 

But that was not enough in this home to 1.45 Bn people with 65% living in villages, where 22 recognised languages are spoken in more than 19,500 dialects. “We felt the need to create not necessarily a doctor, but kind of a junior doctor or someone who could help patients understand where things were going wrong,” Jain said. 

According to him, what often goes unaddressed by most of the nearly 13,000 healthtech startups is the vast potential in Tier-II and Tier-III towns. A large segment of the population in rural and semi-urban areas remains poorly served by mainstream digital healthcare platforms.

Some have started working on this aspect. Practo and mfine, for instance, are integrating multilingual support and telemedicine services that cater to rural and semi-urban users. Startups like Dozee are working on voice-enabled AI assistants to improve accessibility. 

“You can find many chatbots out there, but a lot of times, the general public, particularly in Tier-II and Tier-III cities, won’t be using ChatGPT or interacting with chatbots. You need something more approachable, accessible, and communicable, because the patients may not be comfortable in English. Expecting them to use your product and get value from it without addressing that is unrealistic,” Jain said. 

Vandit Jain nestled his dream venture around this open space until Bharosa AI was rolled out with Dr Amit Gupta and Nidhi Bharani as cofounders in 2023. 

The founders designed their AI-powered health assistant Mira to simplify the healthcare journey for both the doctor and the patient. The Gurugram-based startup runs everything through Mira. This voice-based medical assistant works over a simple phone call and speaks in over 100 languages and dialects. 

Something More Human, Not Another Chatbot

Like most startups, AI unleashed an opportunity that the founders wanted to tap into, when Jain and Gupta conjured the thought. 

“Tools like ChatGPT were just beginning to show their potential, and it was becoming clearer that AI could play a huge role in improving how people access information, including healthcare,” Jain said.

Bharani, backed by her experience in machine learning (ML) and voice AI, helped the duo spot the gaps in the system and map their way forward, leveraging AI. They were clear in their plan to create something more human than making just another chatbot. 

The founders bought the domain Bharosa.ai in 2024. 

Building Voice-First AI That’s Simpler, Easier, Smarter

The founders spoke with over 100 patients to address the problem at the grassroot level. As a first step, they tried to resolve the problem by helping patients find the right doctor. 

“Our initial idea was to create a review and rating system that would recommend doctors based on specific criteria. We called it the Bharosa Score. It was driven by ML and based on six core pillars identified by Gupta, drawn from his years of medical experience. These included things like quality, availability, hospital resources, and affordability,” Jain said.

Soon they realised that it needed something more approachable. “There are tonnes of chatbots out there, but if you go to Tier-II or III cities, you won’t find people using GPT or typing into bots. People wanted something that feels more natural – a voice that would speak the language they are comfortable in.” 

It is this insight that shifted their focus to building an AI agent (they later named it Mira) that could speak directly to patients, understand their symptoms, and route them intelligently to the right specialists.

Jain said Mira was born out of a simple belief of catering to the average Indian, “All of us believed in the power of AI. Tools like ChatGPT were game-changers, but to truly reach the average Indian, people like you, me, our parents, we knew it had to be simpler. ChatGPT isn’t something my mom can use for a diagnosis. But if I put that same intelligence into a phone call-based health app, she can.”

Bharosa AI claims to be the first in the world to use OpenAI’s Model Context Protocol (MCP) in a real-world healthcare product. With MCP, Mira can retain patient history, doctor preferences, and clinical logic across calls, enabling context-aware, intelligent interactions. 

Designed to assist both patients and doctors, Mira qualifies patient queries, routes them to appropriate doctors, and provides doctors with structured transcripts and summaries. This streamlines the outpatient workflow and reduces the administrative burden.

A Look Into The Tech Stack Behind Bharosa 

Bharosa’s models, based on the OpenAI platform, are fine-tuned and fitted with Function Calling and MCP to remove or mask all personal identifiable information such as clinical transcripts, doctor-patient interactions, structured treatment protocols, diagnostic guidelines, and synthetic datasets validated by domain experts. These inputs ensure that Mira delivers medically coherent and culturally localised responses. The brand is also working towards integration with India’s UHI and global standards like HL7.

Its tech stack is built on containerised microservices. The platform uses multiple technologies such as agentic, retrieval-augmented generation (RAG), prompt chaining, and OpenAI’s Function Calling and MCP to maintain continuity across user interactions. “We store patient history, doctor preferences, and triage data using PostgreSQL and Redis. Our backend, built with Django and FastAPI, processes this data and carefully shares only what’s needed with external AI models,” Jain said.

Bharosa does not train any foundation model of its own, but it does fine-tune OpenAI’s models to suit the needs of the product. The company adopts an MCP and function-calling-oriented Agentic approach, where the base model acts as a triage agent between multiple experts and the user. To fetch further details and context, it can also fetch data through various configured MCP servers.

Bharosa customises models for its enterprise customers, who want to provide MIRA Chat in offline configuration or on air-gapped networks. These include capabilities like multi-modality models based on LLAMA-3.2-Vision-11B for text and image data, along with the Whisper family to transcribe audio. 

On the data security and privacy front, Bharosa follows a defence-in-depth strategy that protects data both in transit and at rest. All data transmitted between users and its systems is encrypted using TLS 1.3 over HTTP/2. Data at rest is secured with AES-256 encryption, controlled by Customer-Managed Encryption Keys (CMEK). Passwords and other sensitive secrets are hashed using PGCrypto, and temporary data is automatically deleted once its purpose is fulfilled.

“When a user places a call through the telephone network, the audio is routed to Bharosa’s backend over secure internet channels for real-time processing. Metadata such as contact information and timestamps is retained for up to 90 days. Call recordings are deleted within 24 hours, unless needed for an ongoing consultation,” Bharani told Inc42. 

This short retention period aligns with the regulatory expectations such as 30-day data retention guidelines, and ensures that the data is kept only as long as necessary. “The recordings may be temporarily extended for cases like doctor referrals, but are removed after use.” 

Bharosa is working on its SOC 2 Type 2 and HIPAA compliance due to their relevance in healthcare and SaaS applicability. It is also aligned with Indian regulations, including the IT Act, 2000, and the Digital Personal Data Protection Act, 2023.

A Game-Changer In The Making?

While the startup is in its pre-revenue stage, it is testing a few monetisation models in the market. “One stream is a lead-based model for doctors, where they pay $1-2 per qualified lead once they see value in the platform. Another key revenue opportunity lies in our AI assistant for clinics, which acts as a virtual receptionist, taking notes, assisting in diagnosis, and streamlining operations. For this, clinics are willing to pay $100 to $200 a month, irrespective of the number of leads. The biggest revenue potential will come from partnerships with hospitals, which we expect to scale around mid-2026 once we’ve built a solid presence,” Jain said. 

The founder added that these hospital collaborations could rake in $5,000 to $10,000 a month per institution. 

The startup is now reaching out to young doctors typically with 10-15 years of experience who are more eager to adopt AI. “There are some specialisations we have identified like pediatrics, surgery and neurology, where AI is needed more.”

Bharosa is offering 15-day trials to clinics as it is refining its models. “While it is difficult to make projections for revenue, but on a conservative side, in the next few months, the brand is looking to expand its footprint in Gurugram and Delhi,” Jain said. 

After Diwali, it plans to enter Bengaluru and Hyderabad, where it is in discussion with leading hospital chains. It is looking to onboard over 300 clinics across Gurugram, Bengaluru and Hyderabad. 

The founder claimed that Bharosa was the first to deploy real-time voice AI over phone calls in healthcare, distinguishing it from app-based models. Its core differentiator lies in building a unified ecosystem powered by its AI agent, Mira, which enables all services to talk to each other. 

Bharosa will have to navigate not only the challenges of brand-building but also the rising competition in India’s healthtech space. The startup has entered a domain where brands are offering AI-driven solutions. 

Practo, for instance, leverages AI for operational optimisation, intelligent patient routing, and personalised healthcare experiences, while Dozee uses AI to analyse data from contactless sensors for continuous patient monitoring, enabling early detection of clinical deterioration. Jivi AI offers a voice-based health assistant that conducts triage and symptom checks through conversational interactions, and Qure.ai focusses on radiology, using AI to interpret chest X-rays and head CT scans.

Beyond market competition, the startup will also need to ensure that its technology is both reliable and effective. Data privacy and security will continue to be a key concern, given the highly sensitive nature of health information deals with. There is also a risk of bias in training data, which can lead to unequal outcomes, especially in a diverse country like India. 

These challenges call for deeper integration between doctors, policymakers, and patients to make sure AI truly serves people with adequate policy support and reliability.

[Edited By Kumar Chatterjee]