Swapnil Chavan

Software Engineer with 2 years of experience building AI-powered backend systems for healthcare, B2B, and SaaS platforms. Specializes in end-to-end LLM systems, including voice AI pipelines, RAG architectures, and LLM-driven automation, alongside scalable microservices using Python, TypeScript, and cloud infrastructure (AWS, PostgreSQL, Kafka), with a proven track record of cutting inference costs by 5x.

Founding Software Engineer at Onicares AI (2026-05 – Present)

• Built DIDI bot, a voice-to-voice maternal-health bot (VAD → STT → LLM → TTS) that assesses symptom severity via a decision-tree mechanism — the LLM selects the right symptom dataset from many available datasets and asks questions sourced from it — registers complaints, and flags emergencies (ambulance vs. home care) at $0.021 per turn, while dynamically switching between Complaint, Assistant, and Mental Health modes with mode-switching handled entirely by the user intent.
• Built Didi Plus, a doctor-facing diagnosis recording bot using an STT + LLM pipeline, with a response latency of 2-3 sec, to record, show, and update diagnosis cards, and dynamically invoke new card templates based on doctor input.
• Built a Sarvam AI + Gemini-powered OCR pipeline to extract and structure patients' clinical test reports into structured JSON output, optimizing per-unit inference cost to $0.04/document and $0.01/additional page.
• Built the end-to-end autopay payment system on Razorpay, designing the schema for provider-agnostic switching so payment providers can be swapped without a rewrite, and handling webhooks for subscription autopay charges; also owned scheduled cron jobs, monitoring, and bug fixes across the platform.

Software Engineer at Klubit AI (2025-08 – 2026-01)

• Built an end-to-end backend using NestJS, TypeScript, PostgreSQL, and Kafka, delivering 9+ scalable modules and high-performance REST APIs with 99.99% uptime, supporting posts, courses, events, and social features.
• Architected and deployed containerized applications on AWS (EC2, S3, ECR) using Docker, implementing S3 pre-signed URLs for secure file storage, automated health checks, and monitoring to maintain 99.9% production uptime.
• Developed a responsive TypeScript-based web application with modern UI/UX and efficient state management, integrating Google APIs and Razorpay (Autopay Marketplace) to streamline authentication, payments, and user workflows.

Software Engineer at Sharbo AI (2024-11 – 2025-07)

• Built a scalable B2B visitor intelligence platform (NestJS, TypeScript) enabling real-time visitor tracking, IP-level enrichment, AI-driven company profiling, and automated lead generation across a microservices architecture with 20+ modules (PostgreSQL, AWS, Cloudflare Queues, Clerk).
• Designed the AI-driven company profiling pipeline, using LLM-based enrichment to turn raw IP/firmographic signals into structured company insights, automatically scoring and qualifying leads against ICP criteria without manual review.
• Integrated LLM-powered summarization and classification workflows into the lead pipeline to convert unstructured visitor and company data into actionable sales-ready insights, reducing manual triage for the sales team.

B.E. in Computer Engineering – University of Mumbai (2025)

Diploma in Computer Technology – Government Polytechnic Pen (2019 – 2022)