Build Healthcare AI Platform for India

We are looking for an experienced AI/ML engineer or small AI engineering team in India to design and build the foundation of a new healthcare technology platform focused on the Indian market. The platform will allow users to upload healthcare documents such as: Medical prescriptions Laboratory reports Medical bills Diagnostic reports Doctor notes Other healthcare-related documents The system should process and understand these documents, extract useful medical information, and present it to the user in a clear and understandable way. Over time, we also want to build a population-level healthcare intelligence capability that can identify potential geographic and temporal patterns in diseases or symptoms, using appropriately anonymized and aggregated information. This is an ambitious product and we are looking for someone who can think beyond simply connecting an LLM API. Important: This Is NOT a POC / Throwaway Project We are not looking for a quick demo or disposable proof of concept. The software developed during this engagement should form the foundation of the eventual product. The architecture, database, APIs, code, AI components, security model and infrastructure should therefore be designed for maintainability and future expansion. Another engineering team should be able to take over the codebase and continue development without having to rewrite everything. What We Expect You to Build The initial engagement will focus on creating the first working product foundation. A typical end-to-end flow should look something like: User → Data Entry → Privacy/Anonymization → Document Upload → OCR/Document Processing → Medical Information Extraction → Knowledge Retrieval/AI → User-Friendly Response → Structured Data Storage The exact architecture is intentionally open. We expect you to evaluate and recommend the appropriate technology rather than simply implementing a predetermined solution. AI Architecture One of the important responsibilities will be determining the right AI architecture. We are considering technologies such as: LLMs SLMs RAG Fine-tuned models Document AI OCR Embedding models Vector databases Knowledge bases Traditional ML/statistical models Hybrid architectures You should be able to explain: Why should we use an SLM? Why should we use an LLM? Where does RAG make sense? Where should deterministic rules or traditional ML be used instead of an LLM? We do not want technology chosen simply because it is currently popular. The architecture should consider: Accuracy Cost Latency Privacy Scalability Maintainability Model performance Hallucination risk Ability to update medical knowledge Future migration between models Medical Document Processing The system should eventually support different types of healthcare documents. For the initial implementation, we will provide representative sample/synthetic documents. The system should demonstrate the ability to process documents such as: Prescriptions Lab reports Medical bills Diagnostic reports The developer should design the document-processing pipeline and demonstrate: Document → OCR/Document Understanding → Structured Medical Information The implementation should consider: Scanned documents PDFs Images Different hospital/doctor formats Indian medicine names Medical abbreviations Different document layouts Potentially poor-quality scans Handwritten prescriptions may be considered where technically feasible, but should not be treated as a guaranteed requirement for the initial release. Healthcare Information & User Experience The system should be able to transform extracted information into understandable responses. For example, if a user uploads a prescription or report, the system could explain: What information was found What medicines/tests are mentioned What medical terminology means What questions the user may want to ask their doctor General precautions or things to monitor Reasonable next steps The system should not position itself as a replacement for a doctor or make unsupported definitive medical diagnoses. Healthcare safety should be designed into the architecture. The developer should consider: Uncertainty handling Hallucination prevention Trusted knowledge sources Response validation Medical escalation scenarios Safety rules Appropriate disclaimers Evaluation of AI responses Privacy & Anonymization — Critical Requirement Privacy is a fundamental requirement of this project. We expect anonymization/privacy protection to be considered at the point where data enters the system, rather than added later. The developer must design an appropriate approach for: PII minimization Anonymization/pseudonymization User identifiers Medical information Location information Demographic information Data storage Data access Data retention Data deletion We want to understand exactly: What information is stored? Where is it stored? What information is sent to an AI model? What information is retained? What information can be used for analytics? The architecture should minimize the possibility of identifying an individual from combinations such as: medical condition + location + age + date/time + demographic information. Population-level analytics should use appropriately aggregated/anonymized information. The developer must also ensure that project medical data is not used to train their own models, another client's models, or third-party systems without our explicit written authorization. Population Healthcare Intelligence This will be an important future component of the platform. With appropriate consent, anonymization and aggregation, we want to explore whether the system can identify unusual patterns such as: Increase in particular symptoms Increase in particular diagnoses Geographic clustering Time-based trends Regional healthcare patterns Potential infectious-disease signals For example: If a statistically unusual number of similar medical events appear within a geographic region over a period of time, can the system identify the pattern for further investigation? We do not expect an LLM alone to determine whether an outbreak exists. We expect the developer to consider appropriate: Statistical analysis Time-series analysis Anomaly detection Geographic aggregation ML techniques Epidemiological principles This capability can be developed in a later phase, but the initial database and architecture should not prevent it from being added later. Database Design The developer is responsible for proposing and implementing the database architecture. We expect a clear data model covering areas such as: Users Consent Uploaded documents Extracted medical information Medicines Reports Symptoms/conditions Location information Demographic information Anonymized analytics data AI interactions Audit information The exact schema and database technology are open to recommendation. We expect the developer to explain why the chosen approach is appropriate. Architecture & Backend The developer will be responsible for defining the initial technical architecture, including: Application architecture Backend architecture Database APIs AI/ML services Document processing RAG/knowledge layer Authentication/authorization Privacy controls Logging/auditing Deployment architecture Future scalability We expect architecture diagrams and technical documentation. GitHub & Source Code We will provide a private GitHub repository for the project. All primary development must take place in the client's private repository. The developer should not maintain the only copy of the project in a personal GitHub repository. All source code and project deliverables must be committed to the client-controlled repository throughout the engagement. The repository should contain appropriate: Source code README Setup instructions Configuration documentation Tests API documentation Architecture documentation Deployment instructions Working Example — Mandatory At the end of the engagement, we expect a working end-to-end example, not just architecture diagrams or source code. For example: User uploads a sample medical document Document is processed Relevant information is extracted Information is structured Appropriate knowledge is retrieved AI generates a grounded response Response is presented to the user Appropriate data is stored Privacy/anonymization mechanism is demonstrated We should be able to run the system from the private GitHub repository using the supplied documentation. Code Quality We expect: Clean architecture Maintainable code Proper separation of components Meaningful error handling Automated tests where appropriate Secure handling of credentials No hard-coded secrets Clear configuration management Logging and monitoring considerations Documentation We are optimizing for a product foundation, not a demo. Intellectual Property & Confidentiality All project-specific work created as part of this engagement will be subject to the intellectual-property terms agreed in the contract. This includes, where applicable: Source code Architecture Database design APIs Prompts AI workflows RAG configuration Evaluation frameworks Documentation Project-specific datasets Project-specific models/configurations Other project-specific deliverables The developer must keep project information confidential. The developer must not: Share project information with third parties Publish the project as a portfolio/case study without written permission Reuse our confidential architecture or information for another client Use our healthcare data to train another model Copy or retain confidential medical data after the engagement, except where legally required Any third-party or open-source components should be disclosed appropriately. Limited Competitive Restriction Because this project involves proprietary healthcare AI concepts and architecture, we intend to include a limited six-month post-engagement restriction relating to directly competing projects, subject to applicable law and the final written agreement. The intent is to prevent the developer from using our confidential information, proprietary architecture, datasets, workflows or project-specific intellectual property to develop or assist a directly competing product for six months after the engagement. This is not intended to prevent the developer from working generally in AI/ML or healthcare technology. The exact scope will be agreed in the contract before work begins. Who We Are Looking For We are looking for an experienced AI/ML engineer, AI architect, or small engineering team based in India. Strong experience in several of the following is preferred: LLMs RAG SLMs NLP Document AI OCR Python FastAPI or equivalent Vector databases Embeddings Cloud platforms REST APIs SQL/NoSQL databases Docker CI/CD ML evaluation Healthcare AI experience is a significant advantage. Experience with the Indian healthcare ecosystem, Indian prescriptions, medicines or healthcare documents is highly desirable. What We Value We are looking for someone who: Thinks architecturally Can challenge assumptions Can explain technical decisions clearly Writes maintainable software Understands AI limitations Takes privacy seriously Understands that healthcare requires additional safety considerations Can work independently Is comfortable making technology recommendations Takes ownership of deliverables Can build rather than just advise We are not looking for someone whose main experience is assembling generic ChatGPT/LangChain demos. Initial Engagement Location: India / Remote Engagement: Contract Expected duration: Approximately 8–10 weeks Initial budget: ₹2,00,000–₹3,00,000 Payment: Milestone-based The exact milestone structure can be finalized with the selected candidate. A successful initial engagement may lead to a longer-term development relationship. Suggested Initial Milestones Milestone 1 — Architecture & Data Design System architecture Database design AI architecture recommendation SLM vs LLM vs RAG evaluation Privacy/anonymization architecture Technology stack Development plan Milestone 2 — Core Product Foundation Private GitHub setup Backend Database Document upload Document processing Medical information extraction Initial AI/RAG implementation Milestone 3 — Working End-to-End Product Complete working flow User interaction Grounded AI responses Privacy controls Error handling Initial testing Milestone 4 — Hardening & Handover Tests Documentation Deployment instructions Architecture documentation Known limitations Final working demonstration Complete handover How to Apply Please do not send a generic proposal. Answer the following questions: 1. Architecture Based on the requirements above, would you recommend: LLM + RAG, SLM + RAG, fine-tuned model, or a hybrid architecture? Explain your reasoning. 2. Medical Documents How would you process a scanned Indian prescription from upload through to structured medical information? 3. Privacy How would you anonymize/pseudonymize healthcare data at the point of data entry while still allowing useful future analytics? 4. AI Safety How would you reduce hallucinations and prevent the system from giving unsafe medical advice? 5. Database What database architecture would you initially recommend and why? 6. Population Analytics How would you approach identifying potential geographic disease trends while protecting individual privacy? 7. Previous Work Give us two examples of AI systems you have actually built. For each, explain: Problem Architecture Technologies Your role Scale Biggest technical challenge 8. Production Code Tell us about a project where you built something that was later maintained/developed by another engineering team. 9. Availability How many hours per week can you commit during the first 8–10 weeks? 10. Commercial Please provide: Hourly rate Expected total cost for this engagement Earliest start date Important Please do not use confidential information from previous clients when answering these questions. We are looking for a long-term technical partner, not simply someone to complete a short coding assignment. If you believe a different architecture would be more appropriate than what we have described, tell us why. We value good technical judgment.