SYSTEMSENGINEER//BUILDER

MAYANKSHARMA

Building systems that don't break under pressure.

Open source. AI infrastructure. Backend depth.

DEPLOYED_PROJECTS >SYS_RESUME.PDF
SCROLL_TELEMETRY
ACT 2 // PRIMARY ARTIFACTS

Engineering Showcases

PRODUCTION SYSTEM CORES — INSTANT CLARITY AT A GLANCE.

ENGINEERING REPORT // PROJ-01

METHER OS

Personal AI operating system. Handles WhatsApp messages, Gmail read/write, meeting scheduling, voice queries, and end-of-day summaries via a single orchestration layer.

PERSONAL COMMITS65+
WORKFLOW ACTIONSWhatsApp / Gmail
SYSTEM ARCHITECTURE MODELWHATSAPP / VOICEGMAIL ENGINEMETHER CORE / GROQ APIOrchestrator & Memory CoreAUTOMATIONSMEETING AGENT
INTEGRATION DEPENDENCIES
FastAPINode.jsReactGroq APIGmail APIWhatsApp Cloud API
CRITICAL ENGINEERING DECISIONS
▲ [DEC-1] Context Isolation EngineDesigned specialized isolation buffers that store task execution context separately from active real-time message streams, preventing memory state contamination.
▲ [DEC-2] Latency Optimized Route ModelImplemented light classification prompts via Groq API instead of multi-step agent chains, reducing intent parsing latency from 4.2s to 650ms.
ENGINEERING REPORT // PROJ-02

AI Dataset Quality Inspector

A diagnostic auditing system that validates uploaded datasets using Pandas and SciPy distribution checks.

SCHEMA PREVIEW45s → 3s
STATISTICAL ENGINES5 Audits
SYSTEM ARCHITECTURE MODELDATASET INTAKECSV / JSONFASTAPI AUDIT1. PSI & KS Tests2. Bias & Fairness3. Distribution DriftSCHEMA AUDIT PANELLatency: 45s → 3s
INTEGRATION DEPENDENCIES
FastAPIReactPandasSciPyTailwind CSS
CRITICAL ENGINEERING DECISIONS
▲ [DEC-1] Chunk-Level Statistical StreamingEngineered segmented chunk-based reading for large CSV payloads. Bypasses standard in-memory Pandas loading, preventing out-of-memory container crashes.
▲ [DEC-2] Fairness & Bias Audit PipelinesWrote custom SciPy validation routines checking PSI (Population Stability Index) and KS tests for demographic group comparison audits.
ENGINEERING REPORT // PROJ-03

AI Profile Roaster

A Flask and Groq API platform that delivers structured profile analyses, optimized for high concurrency.

GROQ CONCURRENCYOptimized
API FETCHJSON Validation
SYSTEM ARCHITECTURE MODELGITHUB PROFILEUsername InputFLASK CORE + GROQBackpressure QueueSemantic Profile CacheGROQ ROASTHigh Concurrency
INTEGRATION DEPENDENCIES
FlaskPythonGroq APIViteTailwind CSS
CRITICAL ENGINEERING DECISIONS
▲ [DEC-1] Backpressure Task QueueStructured simple batch throttling queues in Flask to restrict concurrent outgoing requests to Groq, staying safely below free-tier token barriers.
▲ [DEC-2] Semantic CachingCreated local Redis-style key/value maps of previously queried GitHub profiles, preventing redundant external API calls during viral request spikes.
ENGINEERING REPORT // PROJ-04

GitHub Repo Analyzer

A repository intelligence platform that structures repository file dependencies into interactive network nodes.

DEPENDENCY RESOLVERInstant
PARSING SPEEDRealtime
SYSTEM ARCHITECTURE MODELREPOSITORY URLGit HookVITE + VIZ ENGINENode structural treeGraphQL Fetch QueueREPO INSIGHTSStructural Map
INTEGRATION DEPENDENCIES
ReactTypeScriptViteTailwind CSSGitHub Graph API
CRITICAL ENGINEERING DECISIONS
▲ [DEC-1] Tree Parsing TraversalBuilt recursive dependency structure mapping that extracts import paths directly from packages, displaying them as canvas nodes using custom physics hooks.
▲ [DEC-2] GraphQL Query StaggeringConfigured sliding-window request queues to fetch large repository trees sequentially, preventing standard rate-limit blocks from GitHub.
ENGINEERING REPORT // PROJ-05

LifeLink

A Solidity and blockchain-powered relief allocation platform mapping absolute ledger trust to disaster victims.

GAS FEE REDUCTION-40%
TRANSACTION SYNCRealtime
SYSTEM ARCHITECTURE MODELRELIEF FUNDSWeb3 WalletSOLIDITY CONTRACTGas Batch Packing (-40%)MongoDB RPC indexerIMMEDIATE AIDImmutable Audits
INTEGRATION DEPENDENCIES
SolidityWeb3.jsNode.jsExpressMongoDB
CRITICAL ENGINEERING DECISIONS
▲ [DEC-1] Relief Batch PackingStructured custom Solidity arrays to pack multi-recipient relief audits into single transactions, saving up to 40% in gas fees on EVM nodes.
▲ [DEC-2] MongoDB RPC Event IndexerBuilt a robust Node.js event listener that captures smart contract event logs and updates MongoDB indices, avoiding direct RPC node polling loops.
ACT 3 // TECHNICAL CAPABILITIES

Diagnostic Readout

ENVIRONMENT LOGS, INTEGRATION STACK, AND PR CONTRIBUTIONS.

SYSTEM TOPOLOGY // GRAPH PATHWAYS

Ecosystem Topology

JavaScriptTypeScriptPythonCC++RustNode.jsExpressFastAPIFlaskREST APIsWebSocketsReactNext.jsTailwind CSSFramer MotionGSAPThree.jsHTMLCSSMongoDBPostgreSQLRedisGitDockerLinuxPostmanVercelKali LinuxSolidityWeb3.jsEthereumSmart Contracts
TELEMETRY_LOG // NODE: JAVASCRIPTCOORDINATES: X_80 Y_300
CORECAT: LANGUAGE

Asynchronous runtime loops, standard DOM, ES6+ features.

DIAGNOSTIC AUDIT // OPEN SOURCE

sktime / skpro logs

Timeline mapping core contributions to the open-source statistical machine learning environment.

PR #934MERGED

SquaredDistrLoss Bug Report

CATEGORY: BUG REPORT

Identified and reported bug: SquaredDistrLoss using log_pdf instead of pdf in loss formula

PR #917MERGED

Refactor CyclicBoosting Validation

CATEGORY: MAINTENANCE / REFACTOR

Refactored CyclicBoosting to eliminate feature validation code duplication (fixes #916, extracted repeated validation logic into shared method)

PR #833MERGED

Grammar Fix in Validation Messages

CATEGORY: DOCUMENTATION

Fixed grammar in validation error messages: "needs to be" → "must be" across multiple files

PR #813MERGED

Docstring Typo Fix

CATEGORY: DOCUMENTATION

Fixed typo in docstrings: "logartihms" → "logarithms" in skpro distribution classes

AUDIT VERDICT4 PRs MERGED // VERIFIED
// HUMAN ELEMENT
ACT 4 // HUMAN ELEMENT

Systems Builder

CORE MOTIVATIONS AND ACADEMIC PROFILE.

Mayank Sharma Portrait

I believe in learning by building. Most of my growth has come from writing compilers, designing systems from scratch, structuring complex backend workflows, and actively debugging live performance bottlenecks.

Rather than consuming generic tutorials, I prefer reading source files, testing runtime limits, and tackling hard integration challenges. I care about writing production-grade code that is maintainable, typesafe, and performant.

// EDUCATION

Kalvium x JECRC University

B.Tech CS (Software Product Engineering)

PILANI / JAIPUR, RAJASTHAN

// LEADERSHIP

Smart India Hackathon

Team Lead

STRUCTURAL PRODUCT DIRECTION

// TERMINAL END_POINT // ACT 5

OPEN FOR ENGINEERING.

msharma.dev.in@gmail.comGITHUBLINKEDINDOWNLOAD RESUME
SYSTEM CORES: ACTIVE

Ready to power down? Triggers a cinematic standby mode, simulating old-school grid hardware termination.

© 2026 MAYANK SHARMA. ALL RIGHTS RESERVED.

// Open to engineering opportunities.