Solvathon-2026 – THIT

Problem Statements

Problem Statement 1:-Smart Product Recommender for Health Conversations - Users frequently ask health-related questions and often need support identifying the right products—Rx, OTC,supplements, and FMCG. Today, this mapping from conversation to catalog items is inconsistent. We need an intelligent system that interprets health conversations and recommends 5–10 clinically relevant, safe, available products.

Description:

The goal of this problem is to build a Smart Product Recommender for Health Conversations that sits between the conversation layer and the product catalog. The system should take as input:

* User query text – free-text problem statement and constraints (“I’m diabetic”, “I need supplements for XYZ”, “I need products for XYZ skin/hair issues”, etc.).

* Doctor/assistant response text – advice including conditions, classes, salts, preferred dosage forms, precautions.

* Optional user context – age, gender, pincode, and any known conditions if available.

* Product catalog – including product ID, name, brand, salt/composition, indications/tags, strength, dosage form, pack size, MRP/offer price, ratings (if any), stock and availability by pincode, delivery SLA, Rx/OTC/regulatory flags, etc.

Using this, teams are expected to:

* Understand intent and constraints from the combined conversation: extract symptoms, conditions, drug classes, salts, age group, and explicit constraints (e.g., diabetic, pregnant,allergic to a specific ingredient).

* Generate candidate products by mapping this medical intent to catalog attributes (salt/ingredients, indication, dosage form, special tags like “sugar-free”, “pediatric”), handling synonyms, brand vs generic names, and combination products.

Rank and filter these candidates into 5–10 final recommendations by balancing:

1. Clinical/semantic relevance to the described condition and guidance.

2. Ingredient-level fit (correct salt,strength, etc).

3. Availability and delivery time in the user’s pincode.

4. Price bands and affordability (avoid only ultra-premium or only cheapest).

5. Diversity (avoid showing 10 near-identical SKUs from the same brand/strength).

* Produce short explanations per product that make the recommendation understandable and reviewable, e.g., “Matches suggested salt: pantoprazole 40 mg”, “Sugar-free syrup suitable for diabetics”, “Available for same-day delivery in your pincode”.

Expected Outcomes:

By the end of the hackathon, we expect teams to deliver a working prototype of a conversation-driven recommender:

* Input: user query, doctor/assistant response, pincode, and catalog snapshot.

* Output: a ranked list of 5–10 concrete product recommendations (product IDs from the catalog) with scores and short reasons for each.

* The system should visibly handle nuanced factors like salts, ingredients (e.g., sugar-free), dosage forms, age suitability, and availability/substitutes when the exact product is not present.

* The final products should be clinically and contextually sensible.

Overall, a successful solution will bridge the gap between free-text medical guidance and concrete product choices, so that a user can go from “I have this issue and got this advice” to “here are the 5–10 specific, safe, and suitable products I can buy right now” in a way that is clinically sensible, operationally feasible, and easy to scale.

The final solution can be a service, model, or prototype pipeline that, given a sample conversation and catalog snapshot, outputs the recommended products and reasons in a structured format (e.g., JSON) and can be easily plugged into a chat or consultation flow.

Problem Statement 2 :- A real-time, region-language voice AI assistant that works smoothly, is accurate, and costs < ₹2/min, making it viable for large-scale deployment.

Description

There is currently no affordable, real-time conversational speech model available in major Indian languages (Tamil, Telugu, Kannada, Hindi) that can support natural, continuous, real-time two-way dialogue. Existing solutions rely on costly, proprietary APIs, where STT → LLM → TTS pipelines cost ₹7–8/min for async and ₹15/min for real-time, even in English.

For Indian regional languages, options are limited and fragmented.

We need a single, seamless conversational speech engine—preferably built on a unified transformer or better architecture (as per the team’s choice)—that integrates speech-to-text, language understanding, and text-to-speech into one efficient model.

This must be de-novo/ innovative and may be built using open-source models and datasets (e.g., AI4Bharat, IISc). It should not be built by stitching together off-the-shelf APIs.

Primary End Users

Patients and caregivers interacting with Apollo digitally (website, app, telehealth portal).

Walk-in patients seeking information at hospital reception desks.

Expected Outcome:

The ideal solution will deliver:

a. A de-novo conversational speech model.

* Real-time capability (<300–500 ms latency).

* Unified transformers architecture (joint STT + LLM + TTS or other efficient design).
* Optimized for Tamil, Telugu, Kannada, and Hindi.

b. Low cost

* Target: < ₹2 per minute for real-time voice conversation.

* Must be significantly cheaper than current market offerings.

* Natural, human-like voice generation.

* Ability to handle open-ended patient queries.

* Safe fallback for unanswerable questions (handover to human).

* Plug-and-play module for hospital desk kiosks.

Problem Statement 3 :-Medication error and adverse drug reaction (ADR) reporting across healthcare facilities currently relies on manual, branch-level processes. Although standardized forms are used, the absence of a unified digital system limits seamless data consolidation, real-time interpretation, and organization-wide insights.

Description:

* Medication errors and ADRs are captured through standardized formats that are later manually transferred into digital files.

* Analysis is performed periodically (monthly, quarterly, annually), offering insights but without real-time capabilities.

* Facilities maintain extensive historical data—spanning more than a decade—which presents a valuable opportunity for trend analysis, forecasting, and predictive modelling.

* A centralized digital platform can enhance visibility, streamline reporting, automate alerts, and reduce manual workload.

* Integration with EMR systems can enable continuous data capture and improved support for clinicians, pharmacists, and quality teams.

Expected Outcome:

* A unified, user-friendly digital reporting system adopted across all locations.

* Real-time alerts for high-risk medications, potential drug interactions, and adverse events.

* Analytics-driven insights that support timely corrective actions and enhance patient safety.

* Effective utilization of historical data for predictive modelling and long-term trend forecasting.

* Prediction of:

— Likely occurrence of serious medication errors

— Potential severe drug interactions

— ADR likelihood in specific patient groups

— Future error trends based on prescribing patterns and historical behaviours.

Problem Statement 4:-Radiotherapy departments frequently struggle with daily patient scheduling due to unpredictable combinations of simulations, new patient starts, and ongoing treatments. This results in long waiting times, workflow congestion, inefficient machine utilization, and dissatisfaction among patients and staff. There is a need for a unified digital scheduling system that categorizes patients based on treatment status, allocates appropriate time slots, and predicts daily workload to ensure smoother clinical operations and fair patient management.

Description:

On many days, radiotherapy centres experience scheduling chaos—multiple simulations, new case starts, and a high volume of ongoing treatments all overlapping. Current scheduling is largely manual, leading to delays and long waiting times for regular patients who already have fixed treatment slots. A smart, automated scheduling tool is required to organize daily appointments, reduce patient waiting, and balance clinician and machine workload.

Expected Outcome (Ideal Solution Features)

The ideal solution should include:

1.Automated Categorization

* Simulation cases.

* New patient starts.

* Ongoing RT patients (with fixed slots)

2. Smart Scheduling Engine.

* Assigns appropriate time blocks based on treatment type.

* Minimizes waiting times.

* Ensures fairness for ongoing patients.

3.Real-Time Dashboard.

* Displays list of all patients scheduled for the day.

* Shows expected duration for each process (simulation, setup, treatment, QA, etc.).

* Allows drag-and-drop adjustments for sudden changes.

4. Load Prediction

* Estimates daily workload.

* Warns of overbooking.

* Suggests redistribution across machines or timings.

5. Integration Options

* Can integrate with HIS / EMR.

* Can function independently if integration is not possible.

Data Availability

Anonymized data can be provided, such as:

* Daily treatment rosters.

* Number of simulations per day.

* New patient starts.

* Ongoing patient treatment lists.

* Average time per procedure (simulation, setup, IMRT, VMAT, etc.).

* Historical waiting times.

* Machine occupancy logs.

Problem Statement 5:-Evaluation and development of strategies for enhancement of Emotional Intelligence (EI) of frontline nurses.

Description:

Emotional Intelligence (EI) is the ability to understand and manage emotions—our own and others’. It involves noticing feelings, interpreting them correctly, and using that awareness to guide actions. EI is built on four skills: self-awareness, self-management, social awareness, and relationship management.

In healthcare, EI is not just helpful—it’s essential. Nurses work in emotionally intense, high-pressure settings where staying calm, empathetic, and clear-minded directly impacts patient care. Strong EI helps nurses build trust, communicate well, collaborate effectively, and make better decisions, ultimately creating a safer and more supportive environment for patients.

Research shows that nurses with low EI experience greater stress and poorer well-being, while those with higher EI demonstrate stronger teamwork, compassion, and clinical performance. Studies also highlight that paediatric nurses with higher EI communicate better with children and families and deliver more empathetic, holistic care.

The goal is to improve nurses’ emotional resilience, communication skills, critical thinking and empathy, thereby fostering a positive work environment and better patient outcomes.

Expected Outcome:

Solution/ Method

Assessment of current EI of nurses.
Strategies to enhance EI.

* Training Module for nurses.

* Counselling

* Managers training on positive work climate and team work.

Post- Assessment of EI of nurses.
Sustenance plan.

Emotional Intelligence of nurses could be assessed using various standardized tools such as:

Trait Emotional Intelligence Questionnaire-Short Form (TEIQue-SF) developed by Petrides (2009).
Hall Emotional Intelligence Test (HEIT).
Schutte Self Report EI test (SSEIT).
Nurse emotional intelligence questionnaire developed by the researchers by Naser (2016) and Taha (2017).

Problem Statement 6:-Empower elderly care nurses with AI-driven tools that integrate real-time insights and emotional context, addressing personalized care needs and improving patient outcomes.

Description:

Elderly care nurses face challenges managing increasing patient loads, personalized care needs, and time constraints. Current tools focus on task efficiency but lack adaptive, patient-centric support. There is a critical gap in empowering nurses with AI that integrates real-time insights, lived experiences, and emotional context to enhance elderly care.

Expected Outcome:

* Develop a Clinical Agentic AI assistant designed for elderly care, enabling nurses to access dynamic, empathetic, and data-driven insights.

* Improve care quality, reduce nurse workload, and foster personalized engagement for elderly patients.

* NLP for nuanced patient communication, real-time data synthesis, care prioritization, and cultural adaptability.

* Enhances patient outcomes, nurse satisfaction, and care efficiency in elderly healthcare settings.

Problem Statement 7:-Dynamic queue system to manage the in-unit process of allocating guests to free equipment (especially in a shared resource environment)

Problem Statement

Dynamic Queue Management System – Hospitals currently face challenges in managing patient flow across shared resources (IP, OP, HC) such as X-ray, ECG, ultrasound, treadmill, vital-check and other equipment. Due to manual coordination and lack of real-time visibility of equipment availability, patients experience long waiting time, overcrowding and confusion inside the unit. Floor managers struggle to track queues and allocate guests efficiently, impacting both operational efficiency and patient satisfaction.

Description:

Today, most queue allocation is done manually—floor staff physically direct guests to free stations. This leads to:

* Unpredictable waiting time for patients

* Bottlenecks when multiple guests arrive simultaneously

* Staff dependency to manually entries and check resource availability.

* Crowding in smaller waiting areas.

* Low visibility of status (which station is free, which is occupied, and for how long).

* Poor patient experience, especially during peak hours which is impacting the same day review.

Given the lack of an integrated digital system, managing flow becomes reactive instead of proactive, reducing throughput and slowing down overall turnaround time for health checks.

Expected Outcomes

* A minimum viable product (MVP) that can manage queues for IP, OP, and Health Check (HC), assigning patients dynamically to available equipment.

* Low cost that allows us to scale.

* Ability to integrate with the HIS system.

* Automated communication triggers to notify patients about their turn, reducing confusion and waiting-area crowding.

* A simple dashboard for floor managers to easily monitor the queue, equipment status, and patient flow in real time.

Problem Statement 8:-AI based open source drug interaction checker for medications available in India including allopathy , Ayurveda and Unani medicines.

Description:

Patients use medications which are prescribed by a variety of practitioners including allopathy and other systems of care . There is often a lack of understanding that various medications can interact with each other even across systems of medicine . Unlike allopathic medication interactions there is a lack of awareness and evidence on the nature , extent and effect of such interactions . The use of AI to query open data sources comprehensively and build an interaction checker with listed effects will be very useful to providers and patients . Some drug – drug interactions have been well characterised but others are yet to be discovered . A framework to catalog , detect interactions and ultimately adjust the medication prescription will be an outcome of the hackathon .

Expected Outcome:

The ideal solution will be a natural language searchable interface that can obtain primary source information from open source and proprietary databases and other sources including research publications . The solution will rank the confidence of the recommendation and clearly highlight the references that informed the outcome.

Be Part of India’s Next Healthcare Breakthrough

Every edition of Solvathon has sparked new ideas, new collaborations, and new possibilities. This year, the focus is sharper: impact, scalability, and future-readiness. If your team wants to build something meaningful, this is your chance to create solutions that can touch millions of lives. Register Now

Solvathon 2026 - Where Ideas Heal the Future

Problem Statements

Be Part of India’s Next Healthcare Breakthrough

Timeline

Why You Can't Miss It

Innovation at Real-World Scale

Work with the Best Minds

Mentoring That Accelerates You

Build. Validate. Deploy

We've set the stage

Eligibility Criteria

Please Note

Explore Boundless Opportunities by winning our Solvathon

Judging Criteria

Registration

Applications are now closed. Shortlisted applicants will be contacted as per the next steps.

FAQs

Organizer’s contact Information

Mr. Aashish Kumar Jain

solvathon@transformhealth-it.org

QUICK LINKS

LEGAL