I am Swayam R. Gavankar, a systems and backend engineer specializing in low-level optimization, microservices architecture, and agentic workflows. With a background in multi-year backend development, I excel at moving systems from zero to production while eliminating architectural overhead. I take a highly disciplined approach to development, featuring a structured method of agentic coding where I write strict design blueprints and plans for AI agents to execute. This workflow maximizes efficiency, minimizes hallucinations, and guarantees clean, deterministic codebases. As a quick and adaptive learner, I rapidly master new technologies through documentation. Currently, I have completed my 12th-grade education and am preparing for undergraduate engineering admissions, continuing to build high-performance systems and local AI tooling.
Authored a high-performance network transport and mesh routing ecosystem in Rust built on QUIC. Includes `stric-core` (symmetric P2P wrapper over `quinn`), `stric-tower` (an `axum`-like request-response framework), and `stric-flow` (topic-based mesh routing and state sync engine using Dijkstra calculations).
Designed and engineered in a compressed 2-day hackathon timeline. Built as a local-first CLI utility and compatibility inspector for the HuggingFace + PyTorch + CUDA ecosystem. Features precise VRAM/OOM estimation for LLMs (accounting for KV cache size, sequence length, and quantization) and an automated AI-powered incompatibility reporter using Watsonx Orchestrate.
Engineered a secure, distributed, automatically-versioned hierarchical monorepo workspace. Implemented challenge-response authentication (Ed25519), PASETO session management, local keyring encryption (AES-256-GCM, Argon2id), custom folder access control bitmasks, and hierarchical cache invalidation via Redis.
Created a local-first, privacy-focused desktop assistant using local SLMs. Built a multi-agent orchestrator utilizing specialized prompts and model adapters (Policy, Academic, Mental Health, Fusion agents), local client state encryption (AES-128-CBC + HMAC-SHA256), and Next.js + MUI frontend.
Engineered a small but efficient proprietary options trading microservice in Python using Pandas, NumPy, and FastAPI to process trading orders and option chains in milliseconds. System processes millions of trading symbols and 1000s of options concurrently, executes custom business logic, manages real-time active trade states, and handles automatic exit signals via trading APIs.
Built a server-side rendered blogging platform using Django and Django Rest Framework. Programmed an asynchronous, client-side commenting and reply system leveraging browser fetch APIs and CSRF validation, integrated with locally-bundled Summernote WYSIWYG editor.