Founding Engineer @ ServerHive • Systems & Backend Architect • Top 150 GitHub India

Architecting High-Throughput Systems
& Distributed Platforms

Designing low-latency gRPC gateways, Linux sandboxing fabrics, and autonomous agent orchestration loops. Focused on systems-level performance, reliable state persistence, and scaling production fleets to billions of actions.

speedgRPC Latency
Sub-millisecond CAS lookup
1.8ms
memoryFleet Volume
Production actions handled
3B+
starAgent Impact
Cline (69k★) contributor
69K+ ★
Subsystem:
cloud_sync

RBE gRPC Proxy & Cache Engine

High-Throughput gRPC & CAS Gateway

Live Telemetry
Latency
1.8ms
Throughput
10 Gbps / Streamed
Uptime
99.99%
Real-time Kernel Stream:Golang engine
›[grpc-cas] ActionCache lookup: composite indexKey{hash, isAC} resolved collision
›[verify-ac] Merkle tree directory traversal: Tree.Root & Children validated in CAS
›[peer-mesh] 3-node stateful cluster synced; streaming 184MB chunked payload
›[pool-opt] Randomized upstream invocation ID; bypass Cloudflare 100MB body limit
Architecturechevron_right
Interactive Terminal

Explore My Work from the Command Line

Simulating Shravan's Linux edge environment. Query career milestones, systems architecture, or upstream contributions.

shravan@serverhive-edge:~ (zsh)
Quick:

⚡ Welcome to Shravan Vadeghar's Interactive Shell v2.4 (x86_64-serverhive-linux)

Type help to view available commands, or tap any shortcut chip below.

shravan@edge:~$
Career Milestones

Professional Engineering Experience

Scroll down to auto-reveal deep architectural accomplishments, zero-trust infrastructure, and autonomous multi-agent pipelines.

Jul 2024 — Present
dns

Founding Engineer & Platform Architect

ServerHive Developer Cloud Ecosystem

Architected and scaled a complete developer lifecycle cloud platform — spanning autonomous AI node matching (Gemini price-bot), zero-touch workspace provisioning, unprivileged Linux sandboxing, distributed compilation, and 80TB+ distributed artifact storage.

1,800+ UsersActive Architecture

Key Deliverables

Took full end-to-end ownership from initial systems architecture to scaling to 1,800+ developers, 100+ DAU, and 3B+ production build actions.

Architected an autonomous Gemini AI agent (price-bot) and multi-currency billing engine for dynamic workload-to-node matching and budget optimization.

Engineered the core automation and API gateway in Python, Node.js/TypeScript, Go, and Redis handling instant user provisioning, in-browser VS Code, dynamic Caddy TLS, and hypervisor lifecycle.

Designed distributed state synchronization and lease-based scheduling utilizing Redis, PostgreSQL, cgroups v2 CFS quotas, and CPUSets.

Developed custom Linux PAM security extensions in Go/Python to enforce WebAuthn/FIDO2 Passkey MFA at the SSH terminal level.

Built ServerHive Drive: high-throughput distributed storage in Rust (Axum/Tokio) hosting 80TB+ build artifacts with chunked streaming and Anycast routing.

1,800+ Users
Developer Fleet
3B+ Actions
Build Actions
80TB+ in Rust
Artifact Storage
25x vs Hyperscalers
Cost Reduction
Node.js / TypeScript
Python
Golang
Rust
Redis
PostgreSQL
Linux Cgroups
Bubblewrap
PAM / FIDO2
Docker
Jul 2025 — Sep 2025
smart_toy

AI Engineer Intern

TurboML

Architected a production multi-agent AI execution engine, orchestrating advanced tool-calling, state persistence, and NLP pipelines to automate real-time e-commerce and Swiggy delivery workflows over WhatsApp.

< 50ms Tool LoopsView Architecture

Key Deliverables

Designed autonomous multi-agent task execution flows and sub-50ms tool-calling loops for real-time delivery scheduling and partner API integration via WhatsApp.

Built resilient state machines and memory persistence layers to coordinate sub-agents and ensure automated failure recovery across flaky external services.

Engineered low-latency application backends in Node.js, Python, Kotlin, and Rust, synchronizing live transaction states through a lightweight Matrix Bridge.

Ensured thread safety and zero-message loss under burst webhook traffic with queueing and asynchronous worker pools.

< 50ms Tool Loops
Execution SLA
Node.js, Kotlin & Rust
Core Backends
Matrix & Redis
State Sync
Multi-Agent AI
Tool-Calling Loops
Node.js / TypeScript
Python
Kotlin
Rust
Redis
WhatsApp API
Nov 2024 — May 2025
neurology

Software Developer Intern

DigiQuanta Services LLP

Orchestrated conversational and multimodal AI pipelines in Python and FastAPI to autonomously conduct, transcribe, and analyze clinical hospital patient interviews.

+40% ImprovementView Architecture

Key Deliverables

Integrated fine-tuned LLMs and TTS/STT models with specialized Telugu dialect adaptation, improving localized medical speech recognition accuracy by 40%.

Engineered low-latency FastAPI backend endpoints and prompt-chaining pipelines to collect, transcribe, and structure critical diagnostic feedback in real time.

Designed robust real-time audio pipeline handling noise reduction, automated voice activity detection (VAD), and patient sentiment classification.

+40% Improvement
Speech Accuracy
Clinical Patient Intake
Domain
Telugu Dialect Adaptation
Language
Python
FastAPI
LLMs
TTS / STT
Telugu NLP
Audio Processing
PostgreSQL
Apr 2023 — Sep 2023
security

Software Developer Intern

Research Center Imarat, DRDO

Designed and deployed a secure, role-based internal communications dashboard featuring immutable real-time audit logging for defence personnel.

DRDO / Min. of DefenceView Architecture

Key Deliverables

Developed secure inter-office communication platform using Python and Django, featuring strict role-based access control (RBAC).

Built automated Docker and Jenkins CI/CD deployment pipelines on hosted Oracle Cloud infrastructure.

Engineered real-time audit logging with tamper-evident audit trails for security compliance.

DRDO / Min. of Defence
Organization
Oracle Cloud (OCI)
Cloud Infra
Docker & Jenkins
CI/CD
Python
Django
Oracle Cloud (OCI)
Docker
Jenkins
PostgreSQL
RBAC
Audit Logging
End of Career Milestones • 4 of 4 Examined
Architectural Showcase

Systems & Autonomous AI Architecture

Production systems, virtualization daemons, distributed cache accelerators, and upstream open-source contributions.

Developer Lifecycle EcosystemComplete Ecosystem (3B+ Actions)

ServerHive Developer Cloud Ecosystem

Full-stack bare-metal cloud platform covering the entire developer lifecycle — integrating autonomous AI node selection (price-bot), zero-touch workspace provisioning (server-rental-assistant), in-browser VS Code, unprivileged sandboxing, distributed compilation, and automated artifact delivery.

Autonomous Gemini AI agent (price-bot) matching developer workloads to bare-metal nodes with real-time dynamic pricing.
Zero-touch provisioning of isolated user workspaces with per-tenant browser VS Code (code-server) and dynamic Caddy TLS.
Unprivileged Bubblewrap sandboxing with Linux CPUSets, SMT-isolated CPU core scaling, and MemoryHigh throttling.
Real-time Protobuf build_progress.pb stream parser, adaptive process polling, and automated OTA artifact discovery/delivery.
PythonGolangRustBubblewrapCaddyRedisPostgreSQLTelethongRPC
AOSP Platform & CompilersAOSP Core Platform (C++ & Linux)

AOSP Platform Build Engine & Custom Ninja Compiler

Large-scale Android Open Source Project (AOSP) compilation engine featuring a custom-patched upstream C++ Ninja compiler, Soong anti-starvation scheduling, dynamic Redis concurrency limits, and Bubblewrap sandboxing.

Patched Google's upstream external/ninja (C++20) for AOSP, implementing Node::ForEachOutEdge for high-speed DAG traversal across 180,000+ build graph nodes.
Engineered an anti-starvation scheduler allowing local generator jobs (Soong/Blueprint graph generators) to bypass concurrency caps, preventing remote REAPI worker starvation.
Redis-backed dynamic QoS: single MGET queries fetch user parallelism, tier, and local load average in a single roundtrip with robust RESP parsing.
Multi-tenant isolation: Bubblewrap unprivileged user namespaces, cgroups v2 CFS quotas, and PSI memory pressure throttling for multi-hundred GB AOSP trees.
Telemetry and user resolution: dynamically maps BUILD_USERNAME from build paths, per-tenant log paths, and automated OTA cloud artifact delivery.
C++20NinjaAOSP (Android 14/15)Soong / BlueprintRedisBubblewrapLinux Cgroups v2
High-Throughput gRPC & SystemsFlagship Gateway (Go & Rust)

RBE gRPC Debug Proxy & Distributed Cache Gateway

High-performance federated Remote Execution API (REAPI v2) gRPC gateway and caching proxy in Go and Rust. Eliminates hash collisions, traverses Merkle trees for cache integrity, and orchestrates a stateful 3-node cluster mesh.

Introduced composite indexKey{hash, isAC} in BlobDBCache to eliminate critical cache collisions between Action proto and ActionResult.
Built --verify_ac_blobs: recursively traverses Merkle tree directory hierarchies (Tree.Root & Children) to guarantee referenced CAS output blobs exist before returning hits.
Engineered stateful 3-node peer mesh cluster with peer discovery, distributed FindMissingBlobs, and automated dangling cache eviction.
Dynamic connection pool optimization: randomizes upstream invocation IDs every 30-60s, and streams large uploads directly to bypass Cloudflare 100MB limits.
Dual-engine architecture in Go (RocksDB) and Rust (Sled/Tokio/jemallocator) delivering 1.8ms sub-millisecond CAS lookup latencies.
GolangRustgRPCProtobufRocksDBSledNativeLinkBazel REAPIValkey
Systems & CloudRust Systems

ServerHive Drive

Custom high-throughput distributed cloud storage daemon written in Rust, delivering chunked streaming uploads and encrypted user-quota management.

Engineered with Tokio, Axum, and Hyper for asynchronous non-blocking I/O at scale.
Chunked streaming proxying to backend object stores with custom token and hash verification.
In-memory DashMap route caches and signed cookie verification for zero-latency auth.
RustAxumTokioTowerHyperDashMapReqwest
Systems & CloudZero-Trust Security

Zero-Trust SSH Bastion Gateway

Security-hardened SSH proxy plugin in Go to intercept, inspect, and dynamically route SSH sessions into private LXC containers with WebAuthn/FIDO2 MFA step-up.

Interactive WebAuthn passkey step-up authentication directly over SSH keyboard-interactive prompts.
Dynamic LXC container IP resolution and in-memory positive/negative route decision caching.
Structured audit logging emitting security-sanitized JSON telemetry without credential leakage.
GolangWebAuthn / FIDO2LXCSSH ProtocolLinux PAMPostgreSQL
Systems & CloudHypervisor Daemon

Autonomous KVM Virtualization Engine

Autonomous Telegram bot and gRPC daemon running on bare-metal KVM hosts to control VM lifecycle, snapshots, and automated cloud-init OS reinstalls.

Talks directly to libvirtd over local Unix sockets for sub-millisecond hypervisor control.
Automated OS reinstallation pipeline supporting Ubuntu, Debian, Alpine, Arch, and Windows.
Exposes gRPC server (port 50051) for external automated provisioning from the ServerHive fleet.
PythonlibvirtKVM / QEMUgRPCCloud-InitProtobuf
Systems & CloudBuild Cache Optimization

Bazel Remote Cache & RocksDB Engine

Upstream enhancements for bazel-remote remote build cache server, introducing RocksDB small-blob tiered caching and compact binary metadata.

Implemented RocksDB-backed tier for blobs <512KB to minimize filesystem inode overhead.
Designed 24-byte compact binary metadata representation with Greedy Dual-Size Frequency (GDSF) eviction.
Zero-byte fast path optimization reducing RPC round trips during remote build execution.
GolangRocksDBBazelgRPCCgoREST
AI & AgentsProduction AI Agent (< 50ms)

TurboML Multi-Agent Delivery Engine

Complex multi-agent AI execution engine orchestrating real-time tool-calling, recommendation, and NLP pipelines to automate Swiggy e-commerce deliveries over WhatsApp.

Sub-50ms tool-calling execution loop connecting autonomous planning agents to live delivery APIs.
Resilient state machines and persistent memory handling sub-agent coordination and automatic failure recovery.
Core low-latency backend built in Node.js, Kotlin, and Rust communicating through a Matrix Bridge.
Thread-safe architecture under burst webhook traffic with zero message drop.
Node.js / TypeScriptPythonKotlinRustMulti-Agent AITool-CallingRedisMatrix Protocol
Internal Repo
AI & AgentsHealthcare AI

DigiQuanta Conversational Clinical AI

Conversational AI agents autonomously conducting, transcribing, and analyzing hospital patient interviews with specialized Telugu dialect recognition.

Fine-tuned LLMs and TTS models for Telugu regional dialects, achieving +40% accuracy over baseline speech models.
Automated clinical questionnaire pipeline extracting symptoms and urgency scores for medical staff.
PythonFastAPIFine-Tuned LLMsTTS / STTTelugu NLP
Internal Repo
AI & AgentsML & Data Science

Distributed Build Time Predictor

End-to-end Machine Learning pipeline predicting job execution times on cloud clusters using multi-gigabyte build telemetry.

Trained PyTorch LSTM neural networks and Random Forest regressors on gigabytes of historical build logs.
Feature engineering and hyperparameter tuning to accurately predict node throttling and completion ETA.
PythonPyTorchLSTMScikit-LearnPandasNumPy
Open Source715+ Stars

Playlist AutoUpdater

Auto-update M3U playlists daily without needing an app, running zero-maintenance automation via GitHub Actions. Widely adopted open-source utility.

Earned 715+ GitHub Stars and 68 forks from global users.
One-click terminal deployment with automated scheduled GitHub Actions workflow execution.
ShellBashGitHub ActionsLinuxcURL
Open SourceMerged Upstream (69k★)

Cline Autonomous Coding Agent

Contributed to Cline (69k+ stars), the leading autonomous coding agent IDE extension. Optimized message parsing to eliminate UI thread hangs.

Authored PR #3425 (Merged): 'fix(assistant-message): optimize parsing to prevent hangs on large text renders'.
Streamlined streaming regex extraction, resolving high CPU utilization on massive token outputs.
TypeScriptReactVS Code Extension APIAutonomous Agents
Code & Docs
69,1627498
Open SourceMerged Upstream

BlackboxProtobuf

Contributed to NCC Group's Blackbox Protobuf suite for decoding Protocol Buffers without .proto definitions.

Authored PR #31: Resolved critical Burp Suite hanging issues when parsing large streams.
Reduced memory footprint by 1.5x through efficient byte chunking.
PythonProtocol BuffersBurp SuiteReverse Engineering
Open Source714+ Stars

Tata Sky IPTV Portal & Stream Engine

Automated direct M3U playlist generator and token reverse-engineering engine for IPTV portals, with live stream resolution and token rotation.

Earned 714+ GitHub Stars and 133 forks across global community users.
Reverse-engineered portal APIs and token generation mechanics for seamless TiviMate and Kodi integration.
PythonReverse EngineeringIPTV / M3UKodiTiviMateAPI
Open SourceCI/CD Pipeline

Windows Subsystem for Android Automation Pipeline

Automated CI/CD pipeline script to build and bake Windows Subsystem for Android (WSA) packages integrated with Google Apps (GApps) in one click.

Eliminates manual environment setup using automated GitHub Actions workflow runners.
Integrated multi-architecture Magisk root injection and OpenGApps bundling.
ShellGitHub ActionsAndroidWSAAutomation
Technical Capabilities

Comprehensive Skills Matrix

Core proficiencies across low-level operating system internals, distributed systems, machine learning pipelines, and autonomous agents.

Backend & Distributed Systems

High-throughput gRPC, Merkle caching, concurrency & state engines

Golang (Go)• gRPC & High-Volume Routing
Rust• Tokio Async, Sled & Memory Safety
Node.js & TypeScript• High-Concurrency APIs
Python (FastAPI/AsyncIO)• Distributed Microservices
gRPC & Protobuf• REAPI v2 & Binary Wire Specs
RocksDB & Sled• Small-Blob Embedded Storage
Redis & Valkey• Lease Locks & Distributed State
PostgreSQL• ACID Transactions & AsyncPG
Merkle Tree Traversal• Action-Tree Cache Verification
Composite Indexing• Cache Collision Elimination

AI & Autonomous Agent Backends

Multi-agent workflows, sub-50ms tool loops & state machines

Multi-Agent Orchestration• Sub-50ms Tool Loops
State Machines & Memory• Automated Failure Recovery
WhatsApp & Chat Webhooks• High-Burst E-Commerce
Matrix Protocol• Distributed Agent Bridge
Agent Loop Engineering• Cline (69k★) Runtime
Multimodal AI Pipelines• Speech STT/TTS & LLMs
Gemini AI Agent• Dynamic Price & Node Matching
Prompt-Chain Evaluators• Regression & Latency Testing

AOSP & Systems Platform Engineering

Custom C++ Ninja compiler, Soong build graphs, sandboxing & kernel QoS

AOSP (Android 14/15)• Platform Architecture & Trees
Upstream Ninja (C++20)• DAG Out-Edge Traversal Patch
Soong & Blueprint• Android.bp Build Optimization
Anti-Starvation Sched• Generator Job Prioritization
Bubblewrap (bwrap)• Unprivileged User Namespaces
Linux Cgroups v2• CFS Bandwidth & CPUSets
Kernel PSI Telemetry• Pressure Stall Throttling
Redis Concurrency QoS• Single-RTT MGET Limits
WSA Automation• Magisk & GApps Integration
Bazel Remote Caching• AOSP REAPI Cluster Tiering

Protocols, Networking & RPCs

Streaming architectures, connection pooling & zero-trust security

Google ByteStream• Chunked Resource Streaming
Connection Pool Tuning• Invocation ID Randomization
WebSockets & SSE• Real-Time Terminal Streaming
RESTful APIs• High-Traffic Stateless Endpoints
WebAuthn / FIDO2• SSH Bastion Passkey Auth
Cloudflare Optimization• Bypassing 100MB Body Limits
Caddy Dynamic TLS• Automated Edge Ingress
HTTP/2 & HTTP/3• Multiplexed Streams

Cloud & Bare-Metal Infrastructure

Kernel sandboxing, hypervisor control & automated deployments

Linux Cgroups v2• CFS Quotas & Memory Throttling
Bubblewrap (bwrap)• Unprivileged User Namespaces
KVM / libvirt• Automated Hypervisor Daemons
Docker & Containerd• Multi-Stage Lightweight Builds
GitHub Actions CI/CD• Matrix Automated Pipelines
Jenkins• Enterprise CI/CD Pipelines
Google Cloud (GCP)• Cloud Storage & Compute
Oracle Cloud (OCI)• Hosted Bare-Metal Compute

Full-Stack & Developer Tooling

High-traffic web interfaces & developer experiences

Next.js 15 App Router• High-Performance SSR
React 19• Modern Component Architecture
TypeScript• End-to-End Type Safety
Tailwind CSS v4• Utility-First Design Tokens
Material Design 3• Dynamic HCT Color System
VS Code Extensions• IDE Tooling & Parsers
Telethon• High-Speed Telegram Client
Competitive DSA• IEEE 1st Prize Algorithms
Community & Advocacy

Open Source & Community Leadership

Ranked among the top GitHub contributors in India, with high-impact upstream patches, global sponsorships, and technical conference speaking.

#150Top Contributor in IndiaRanked on GitHub
5,600+Total Contributions3,500+ in 2025–2026
1,500+GitHub StarsAcross 180+ repos
600+GitHub FollowersGlobal Community

Upstream Open Source Pull Requests

cline/cline★ 69,100+Merged Upstream
fix(assistant-message): optimize parsing to prevent hangs on large text renders (#3274)

Streamlined token output parsing and regex execution in Cline's autonomous coding engine, eliminating UI thread freezing on large generative outputs.

Fix Burp hanging issues during stream inspection

Resolved critical thread lockups in Burp Suite when inspecting raw Protocol Buffer streams without definitions, reducing memory footprint by 1.5x.

microsoft/devhome★ 90+Authored Upstream
feat: Add support for Remote Server CPU & telemetry monitoring

Engineered a C# widget for Windows Dev Home allowing developers to track real-time CPU utilization and metrics from remote Linux servers.

feat(security): bubblewrap user sandboxing feature flag and shell (#114)

Implemented Bubblewrap unprivileged user-space sandboxing with CPUSets and SMT-isolated CPU core scaling across the multi-tenant build fleet.

fix(core): bypass node-pty on WSL for Windows executables (#27354)

Resolved critical terminal emulation errors when launching Windows binaries inside WSL environments, and added custom Gemini base URL support.

Refactor dynamic HMAC token generation and IPTV portal session refresh (#65)

Re-engineered token generation mechanics and live stream resolution, restoring service access across Kodi and TiviMate IPTV players.

feat(google): make Gemini web search base URL configurable (#54116)

Enabled self-hosted proxies and custom Gemini API endpoints for autonomous multi-platform agent search queries.

Technical Speaker

GitHub GitTogether

Invited technical speaker at GitHub GitTogether, presenting architectural deep dives and systems concepts to 100+ industry software engineers and platform architects.

Developer Community

Mentored 500+ Students

Core mentor with Google Developer Student Clubs (GDSC-MVSREC) and collegiate open-source initiatives. Guided 500+ students through Linux internals, cloud architecture, and systems programming.

Awards & Academic Distinctions

IEEE Computer Society

1st Prize — Rapid Code

Fastest algorithmic problem solving and competitive systems coding under strict time constraints.

Department of Information Technology

1st Prize — Project Expo

Best engineered systems software and distributed cloud infrastructure demonstration.

Computer Society of India (CSI)

3rd Place — National Hackathon

Built 'Gamified Learning Platform' in 'Code-The-Verse', competing nationwide across India to deliver an interactive real-time educational system in 36 hours.

Dept. of Information Technology

Graded A++ in Logic Building

Highest academic distinction awarded for algorithmic design and computational logic.

Project-Based Courseworks

10/10 Perfect Grade

Graded a flawless 10/10 across all university project-based courseworks, systems labs, and engineering implementations.

Initiate Handshake

Let's Build High-Throughput Systems Together

Excited to bring my experience in high-performance gRPC backends, sub-50ms autonomous agent loops, and distributed caching to building resilient infrastructure that empowers millions of users.

Direct Email

vadeghar.shravan@gmail.com

Fastest channel for professional inquiries, technical interviews, and architecture discussions.

Telegram Instant

@Shra1V32

Direct Telegram contact for quick chats, ServerHive questions, and real-time collaboration.

Professional Network

Shravan Vadeghar

Connect on LinkedIn for professional endorsements, peer recommendations, and roles.

Code Repository

github.com/Shra1V32

Follow my open-source work, starred research repositories, and active platform development.

Location & Availability

Hyderabad, India

IST (UTC+5:30) • Flexible for Global Syncs
Open to Remote & Relocation
SV
Shravan Vadeghar

Backend & Systems Architect. Founding Engineer at ServerHive & RBE Gateway Architect. Designing high-throughput gRPC services, zero-trust infrastructure, and autonomous AI agent loops.

© 2026 Shravan Vadeghar. Built with Material 3.
Systems Online • Hyderabad, India
Next.js 15 • React 19 • Material Design 3