SOSMed AI — Care Documentation & Session Management Platform
A full-scale platform built for care providers, combining automated attendance tracking with AI-driven session documentation. Designed to cut administrative overhead while improving compliance and record accuracy.
Key features:
- Facial recognition-based attendance tracking for staff and clients
- AI-powered session recording with automatic note-taking and quarterly summary generation
- Custom templating engine for building any documentation form without code changes
- Native iOS/Android mobile apps for on-the-go access
- Admin dashboard with real-time reporting and analytics
Live: https://sosmed.ai/
1
42
A quick look at two core screens from DFY Reels, a content SaaS platform I built end-to-end, frontend to payment infrastructure. Built with React/Next.js and a Node.js/TypeScript backend.
Live: https://app.dfy-reels.com/
1
47
Custom LLM Fine-Tuning — Indian Legal Q&A
A custom LLM fine-tuned using LoRA adapters, purpose-built for Indian legal reasoning — trained on statutes, the Constitution, and real court case data to deliver accurate, explainable answers to legal questions. Deployed on cloud infrastructure running the fine-tuned weights — no third-party LLM API dependency, full data sovereignty, no vendor lock-in. End-to-end build: data pipeline, fine-tuning, iterative retraining, evaluation harness, and a working demo.
Key details:
- LoRA adapter fine-tuning of an open-source reasoning LLM for a specialized legal domain
- Deployed on cloud infra running custom fine-tuned weights — no third-party LLM API dependency, no vendor lock-in
- Data pipeline merges synthetic chain-of-thought examples with real judicial case data and constitutional/statutory text
- Chain-of-thought reasoning for explainable, auditable legal answers
- Iteratively retrained across multiple checkpoints, validated with a baseline-vs-fine-tuned evaluation harness
- Interactive chat demo for walkthroughs
1
35
AI-Powered Healthcare Platform
Built a full-stack AI healthcare SaaS serving 1,000+ active users. Shipped the iOS app to the App Store, engineered the Node.js backend, PostgreSQL database, and AWS infrastructure from scratch. Authentication rebuilt with Firebase Auth and fine-grained RBAC across patient, provider, and admin roles. AI-driven care workflows integrated throughout with a compliant data consent layer. Diagnosed and resolved Apple guideline rejections, resubmitted and approved. Every layer built for compliance, security, and scale.