# Devanshu Sharma > Systems Software Engineer ## Summary I obsessed over accuracy in distributed systems, databases, and concurrency the most in my previous roles. Being able to differentiate between the tiniest details can determine whether a system works or fails at its core. Currently, I am an intern at Clockwork Labs where I am building SpacetimeDB, a distributed, real-time database that integrates the database with the application server and synchronization layers, all built within a single, high-performance real-time environment. Thus, this is another gigantic puzzle I'd love to work on that has plenty of intricacies to keep me entertained. I have also worked on increasing scheduling performance and security at Cal.com and on GPU attestation and evidence pipelines with deterministic verification at Ashiba Research. I have the most respect for code with clear, unambiguous, and explicit invariance. Currently, I am most passionate about real-time infrastructure, distributed databases, backend performance, and systems software that is open source and built with a significant amount of Go code. I also like the details around real-time indexing, search, schema design, correctness, and the kind of frameworks that make complex systems easier to document clearly. ## Location Dehradun, Uttarakhand ## Technical Interests - Real-time infrastructures and distributed databases - Backend performance that optimizes previously inefficient routes - Open-source systems software, especially Go-heavy codebases - Real-time indexing, search, schema design, and correctness - Clear technical documentation for complex systems ## Skills - C++, Go, SQL: Most comfortable close to systems work: Go for backend and infrastructure, C/C++ for lower-level thinking, and SQL for the database-shaped parts of the world. - Distributed systems, REST APIs, concurrency, gRPC: I like building services that are fast, correct, observable, and boring in production in the best possible way. - PostgreSQL, Redis, etcd, MongoDB: Interested in indexing, query paths, coordination, migrations, and all the little decisions that make data systems trustworthy. - Docker, Git, Linux, Bleve, pgx, libpg_query: The usual kit for building, debugging, shipping, and occasionally staring very hard at a terminal until the truth appears. ## Work Experience - Clockwork Labs (Jun 2026 - Present): - Developing SpacetimeDB, a distributed real-time database that combines storage, application logic, and synchronization into one high-performance system - Improving correctness, scalability, and low-latency access for open-source real-time infrastructure used in modern multiplayer and collaborative applications - Creating practical technical writing and tutorials that make distributed systems and real-time database concepts easier for developers to build with - Ashiba Research (May 2026 - Jun 2026): - Built ARC (Attestation Receipt Compiler), a deterministic verification system that turns GPU telemetry, attestation tokens, and deployment logs into explicit trust receipts - Engineered NVIDIA attestation checks with nonce-based freshness guarantees to detect unsupported, mismatched, or unverifiable GPU identity claims - Designed a canonical evidence pipeline across nvidia-smi, NVML, DCGM, OTEL spans, CloudTrail, and NVIDIA EAT JWTs with clear supported, contradicted, unknown, and out-of-scope claim boundaries - Cal.com (Jun 2025 - Oct 2025): - Improved reliability, performance, and security across high-traffic booking flows on Cal.com's core scheduling platform, including a Stored XSS fix for user-upload handling - Replaced O(n²) scheduling logic with sorting, early-exit scans, and two-pointer techniques, improving key production workloads by 25 times - Strengthened conflict detection, availability filtering, cross-locale correctness, and overall code quality while reviewing 200+ PRs in a large TypeScript monorepo ## Education - Bachelor of Technology in Computer Science and Engineering (2022 - 2026): - Graduated with a Bachelor of Technology in Computer Science and Engineering from Graphic Era Hill University, Dehradun - Coursework included Data Structures and Algorithms, Operating Systems, Computer Networks, Database Management Systems, Cloud Computing, Computer Architecture, Object-Oriented Programming, and Web Development - Most interested in systems classes: databases, operating systems, networks, and the parts of software where abstractions meet hardware and latency ## Projects - Distributed Search Engine (Dec 2025 - Feb 2026): - Built a distributed search system in Go indexing 24K+ Wikipedia documents across 8 shards - Achieved 28ms p99 latency at 7K+ QPS through parallel query execution and multi-layer caching - Designed a hybrid ranking pipeline combining BM25 and vector similarity with 384-dimensional embeddings - Implemented an 8-node sharded architecture with etcd-based service discovery and fault-tolerant coordination - Engineered a parallel fan-out / Top-K merge query engine, reducing end-to-end latency by 12x compared with single-node search - Added Redis caching with distributed locking using SetNX to prevent thundering herd behavior under high concurrency - Deployed a 15-service Docker cluster for reproducible, production-like environments - Diffo: Git-like Schema Versioning System (Feb 2026 - Present): - Built a Git-inspired schema versioning system using a commit DAG for branching, merging, and reproducible database history beyond linear migrations - Implemented DFS cycle detection and merge-base computation to preserve correctness under complex histories - Designed a deterministic ancestry engine using BFS with stable ordering, eliminating nondeterministic schema resolution across environments - Developed a three-way merge engine with structured diffs and typed conflicts instead of fragile text merges - Built schema diff and migration generation with topological sorting over foreign-key dependencies - Implemented live schema capture through PostgreSQL catalog introspection and a canonical DDL pipeline ## Open Source - Vitess - vtgate race condition fix (2026): - Fixed a race condition in vtgate where concurrent goroutines accessed PreSessions, ShardSessions, and PostSessions during Rollback/Release flows - Introduced safe snapshot helpers for thread-safe iteration - Validated the fix with Go's race detector - Supabase - frontend performance and navigation fixes (2026): - Resolved a critical frontend performance issue caused by excessive history.replaceState() calls during rapid scrolling - Designed a rate-limited history wrapper with debouncing and deduplication, eliminating browser SecurityError crashes while preserving URL sync and TOC highlighting - Improved docs navigation by adding dropdown indicators for nested sidebar items - Fixed an incorrect API duration example so docs matched supported formats - SpacetimeDB ecosystem (2026): - Contributing to open-source distributed realtime database infrastructure - Working around correctness, low-latency data access, and scalability concerns for multiplayer and realtime applications - Creating developer-facing technical content to make the system easier to understand and adopt ## Technical Writing - Systems I Am Thinking About (2026-07-18): https://devanshusharma.com/reading/systems-i-am-thinking-about - The talk of the trillions. (2026-05-06): https://medium.com/@devanshusharma658/the-talk-of-the-trillions-e5ce2d07c616 - Agent Sandboxes at Scale: A Distributed Systems Design for AI-Driven Development (2026-04-29): https://medium.com/@devanshusharma658/agent-sandboxes-at-scale-a-distributed-systems-design-for-ai-driven-development-3765f4a9460c - I built Git for database schemas (2026-03-30): https://medium.com/@devanshusharma658/i-built-git-for-database-schemas-f259ce3d3cc5 - Building a High-Performance Distributed Search System for 10,000 QPS. (2026-02-09): https://medium.com/@devanshusharma658/building-a-high-performance-distributed-search-system-for-10-000-qps-240e5ce48016 ## Contact - Email: mailto:devanshusharma658@gmail.com - GitHub: https://github.com/Devanshusharma2005 - LinkedIn: https://www.linkedin.com/in/devanshu-sharma-a1baaa243 - Resume: /content/media/resume.pdf ## Important URLs - home: https://devanshusharma.com/ - machine: https://devanshusharma.com/machine - reading: https://devanshusharma.com/reading - projects: https://devanshusharma.com/projects - cv: https://devanshusharma.com/cv - resume: https://devanshusharma.com/content/media/resume.pdf - llms: https://devanshusharma.com/llms.txt - robots: https://devanshusharma.com/robots.txt - sitemap: https://devanshusharma.com/sitemap.xml ## Indexing Preferences Crawlers and AI agents may index, summarize, and use this portfolio for recruiting or professional discovery.