Anubhav Kumar Rao
|
Building systems at scale — from distributed databases to LLM-powered platforms.
Who I Am
I'm a software engineer who thrives at the intersection of systems engineering and applied AI. Currently at Coinbase, I'm working on expanding crypto derivatives markets to the UK and EU while building LLM-powered internal platforms.
Before Coinbase, I worked at D.E. Shaw & Co. on financial time-series infrastructure serving 13M+ tickers and RAG-based semantic search over instruments. I also interned at Rubrik (cloud storage infrastructure) and Razorpay (fintech/payroll).
Outside of work, I build things for fun — a transactional database engine with LSM Trees, a Linux file system simulator, and a real-time Raft consensus visualizer. I also review problems for LeetCode competitions for 5M+ programmers.
I graduated from IIIT Allahabad with a B.Tech in ECE (GPA: 8.65/10).
Education
Indian Institute of Information Technology Allahabad
B.Tech — Electronics & Communication Engineering
Sri Aurobindo Public School
Class 12
4+
Companies
10+
Projects
12TB+
Data Handled
Contest Tester
LeetCode
Where I've Worked
Software Engineer
Coinbase
Remote, India
- Spearheading expansion of Coinbase International Exchange & derivatives to the UK and EU — extending regulated crypto trading to millions of users.
- Built an LLM-powered customer issue aggregation platform ingesting trading, compliance, and verification signals — reducing support contact rate by 15%, saving millions annually.
- Classified customer chat transcripts at scale using LLMs; engineered ETL pipelines cutting project estimation timelines by 50%.
Member Technical (Software Engineer)
D.E. Shaw & Co.
Hyderabad, India
- Integrated RAG-based LLM solution into DESTerm, enabling semantic search across 13M+ financial instruments for quant analysts.
- Architected high-throughput batch processing and log storage for large-scale trading reports, scaling to 2TB with Redis caching for sub-second access.
- Built a time series library powering EOD & intraday data retrieval across 13M+ tickers and 100+ fields — adopted by 10+ teams.
- Engineered quantitative analysis tools (quantile regression, seasonality adjustment) handling up to 12TB of time series data.
Software Engineering Intern
Rubrik
Bangalore, India
- Scaled archival storage backup 4× (8TB → 32TB) for Azure VMs and managed disks.
- Integrated GCP cloud metrics into Rubrik's reporting infra, increasing monitoring coverage by 10% and driving a 17% increase in report traffic.
- Designed a unified reporting, metrics, and licensing framework adopted across all engineering teams.
- Built multi-cloud cost and storage comparison tooling for VMs, Disks, and Databases across Azure, AWS, and GCP.
Software Engineering Intern
Razorpay
Bangalore, India
- Engineered department-level payout filters for XPayroll across 50+ organizations with 100% data migration and zero data loss.
- Designed scalable MySQL schemas, RESTful APIs, and React frontend supporting seamless department onboarding across 50+ enterprise clients.
- Introduced tax deduction schemes for Sections 80DDB, 80TTA, and 80TTB ensuring regulatory compliance across enterprise payroll.
Things I've Built
Personal projects I built to go deep on systems, databases, distributed algorithms, and AI.
A production-grade, self-healing SRE control plane built as a 6-agent swarm (Observer, Diagnoser, Remediator, Safety, Orchestrator, Learner) over NATS JetStream. Monitors 10+ polyglot microservices, performs LLM-assisted root cause analysis with Gemini/OpenAI fallback chain, enforces safety gates with human-in-the-loop approvals via a React dashboard, executes YAML runbook-based remediation, and scores MTTD/MTTR with chaos engineering scenarios. The Learner agent uses RAG over ChromaDB to recommend runbooks from historical incidents.
An interactive, browser-based visualizer for the Hierarchical Navigable Small World (HNSW) algorithm. Simulates multi-layer graph construction, greedy top-down nearest-neighbor search, random-level assignment, and degree-pruned edge linking, with live layer isolation, operation logs, and graph statistics to make vector search internals intuitive.
A from-scratch, educational 32-bit x86 operating system written in C and NASM assembly. Boots via GRUB, runs an interactive shell with syscalls, keyboard input, filesystem, and VGA text-mode display — all with zero external dependencies.
A 5-agent LLM-powered equity analysis system using LangGraph orchestration, integrating RAG over financial reports via Qdrant and real-time sentiment analysis. Delivers structured BUY/SELL/HOLD signals with confidence scores across RSI, MACD, and Bollinger Band indicators.
SQL-compliant database built from scratch with LSM Tree storage, Write-Ahead Logging for crash recovery, and MVCC-based ACID transactions. Achieves 3× faster writes than B-Tree implementations for write-heavy workloads with 10K+ transactions/sec.
Real-time distributed consensus visualizer simulating a 5-node Raft cluster with a virtual networking layer. Enables deterministic fault injection and demonstrates 99.9% consensus achievement under network partitions and node failures.
A real-time collaborative document editor built from scratch using CRDTs (Conflict-free Replicated Data Types). Implements the RGA algorithm with tombstone-based deletion, an insert backlog for out-of-order operations, and deterministic tie-breaking — enabling conflict-free multi-user editing over WebSockets with zero central coordination.
Production-grade file system simulator implementing ext2/ext3 architecture with inode-based storage, directory management, and bitmap allocation. Reduces storage lookup complexity from O(n) to O(1) for 1000+ files.
Coding Profiles & Leadership
Community & Leadership
LeetCode rated contest tester. Problem setter on CodeChef, Codeforces, and HackerEarth. GDSC IIITA wing lead.
Technologies & Tools
Languages
Backend & Systems
AI & ML
Data & Infrastructure
Frontend
Core CS
Deep Technical Writing
Long-form engineering breakdowns with architecture maps, design tradeoffs, incident workflows, and implementation details from projects I build end-to-end.
HNSW: The Algorithm Quietly Deciding What Your LLM Sees as 'Truth'
A practical deep dive into how HNSW powers vector retrieval for RAG systems, why it dominates ANN search, and how poor recall can silently degrade LLM answers.
Total posts
7
Reading minutes
91+
Distinct topics
30
What to expect
- 01Architecture and protocol-level explanations, not just project summaries.
- 02Failure modes, safety constraints, and operator workflow details.
- 03Implementation snippets, design tradeoffs, and test strategy.
Building SRE Agent Swarm: A Multi-Agent Self-Healing Infrastructure Platform
A deep, implementation-level breakdown of how I built SRE Agent Swarm: six cooperating agents, incident FSM orchestration, safety gates, runbook execution, human approvals, and chaos-driven MTTD/MTTR scoring.
Building a Collaborative Editor with CRDTs from Scratch
How I built a real-time collaborative document editor using Conflict-free Replicated Data Types — implementing the RGA algorithm, tombstone deletion, and out-of-order operation handling with zero external CRDT libraries.
Building AnubhavOS: A 32-bit Operating System from Scratch
A deep dive into building a 32-bit x86 operating system from scratch. Features a custom bootloader sequence, memory management, interactive shell, and more.
Let's Connect
I'm always open to interesting conversations — whether it's a new role, a project idea, or just saying hi.
GitHub
See my code & projects
Connect professionally
anubhav100rao@gmail.com
Resume
View full resume (PDF)
Want to work together?
I'm currently at Coinbase but always open to hearing about exciting opportunities.
Send me an email