Copied

Open to remote Mumbai, IN

MANGESH YADAV

AI Automation Engineer / Data Analyst

I design LLM-powered agents, RAG pipelines and end-to-end automation that erase repetitive work — then turn the leftover data into dashboards teams actually use.

See the work Get in touch
Live — inbound call flow ops routed 0000
FIG.01 — AN INBOUND CALL, AUTOMATED END-TO-END IN triggerTrigger — an inbound task or call enters the pipeline STT speech → textSpeech-to-text — the caller's words become tokens RAG vector storeRetrieval-augmented generation — context fetched from a vector store LLM agent coreLLM agent core — reasoning, deciding, calling tools API tools & actionsTools — external APIs and actions the agent can call TTS text → speechText-to-speech — the reply becomes a voice response OUT response shippedOutput — task resolved, zero humans required
Scroll

01/ Experience

Production roles · 2025 → present

Shipped to production.

Aug 2026 — Present

Mumbai, IN

Jolly Clamps Pvt. Ltd.

Automation & Data Analyst

  • Designed and built an internal automation portal that replaced manual, spreadsheet-driven workflows with a single fully automated system — eliminating repetitive daily processing for the entire team.
  • Owned the portal from requirements through deployment — a self-service system that removed the need for daily manual follow-up on core tasks.
  • Converted raw operational data into interactive visual analytics dashboards, giving stakeholders real-time visibility into key metrics and enabling faster, data-driven decisions.
PythonInternal ToolsWorkflow AutomationData Visualization

Sept 2025 — Aug 2026

Mumbai, IN

MCM BPO Pvt. Ltd.

AI Automation Engineer & Python Developer

  • Designed and shipped an AI Receptionist that handles inbound caller inquiries end-to-end — LLM, text-to-speech and speech-to-text integrated into a custom web interface — delivering automated, round-the-clock front-desk coverage and freeing staff for higher-value work.
  • Engineered end-to-end Python automation across core operational processes, cutting manual workflow time by 30% and eliminating repetitive daily tasks for the operations team.
  • Developed an AI-powered content and email engine that generates personalized, multi-platform social posts and bulk outreach campaigns, with built-in engagement tracking.
  • Deployed and maintained production applications on Linux servers — uptime monitoring and incident resolution to keep business-critical automation running reliably.
LLMSTT / TTSPythonLinuxProduction Ops

02/ Projects

Tap to expand

Selected builds.

A full-stack content generation platform built on n8n automation workflows. Type a keyword, and dynamic prompt-engineering pipelines orchestrated across REST APIs produce ready-to-publish posts and images — tuned per platform for Instagram, LinkedIn, Twitter and Facebook.

  • Dynamic prompt pipelines, per platform
  • API orchestration across n8n workflows
  • Image generation built into the flow
Pythonn8nOpenAI APIREST APIsPrompt Engineering
Social Content Engine — n8n workflow visualization
Fig.01 — Workflow orchestration layer

An automated lead-generation tool that fetches and validates high-quality prospects through Apify and Apollo API integrations, submits website forms with Selenium, and sends personalized email outreach over SMTP — end to end, hands off.

  • Lead fetch + validation via Apify & Apollo
  • Automated website form submission
  • Personalized outreach at scale, over SMTP
PythonApify APIApollo APISeleniumSMTP
SalesHunter — automated outreach pipeline
Fig.02 — Outreach pipeline

A full-stack Django video-conferencing web app with server-side token generation and room management. LiveKit WebRTC handles real-time audio/video streaming, with mic and camera controls plus end-to-end encrypted passphrase handling.

  • Server-side token generation & room management
  • LiveKit WebRTC audio/video streaming
  • Encrypted passphrase handling
DjangoLiveKit SDKWebRTCJavaScriptHTML / CSS
LetsMeet — real-time video conferencing platform
Fig.03 — Live session mesh

The Flask backend of a deepfake detection system: a pretrained CNN with cosine-similarity face verification across video frames, plus a modular pipeline for facial alignment, feature extraction and real-time frame-by-frame video analysis built on OpenCV and dlib.

  • Pretrained CNN + cosine-similarity verification
  • Facial alignment & feature extraction pipeline
  • Real-time, frame-by-frame video analysis
PythonFlaskOpenCVdlibCNNDeep Learning
Deepfake Detection — frame analysis pipeline
Fig.04 — Frame analysis pipeline

03/ Stack

pip freeze — daily drivers

The toolbox.

$

Successfully installed mangesh-yadav-2025.1 — 34 skills, 0 bloat

Languages & Frameworks

the daily drivers

PythonJavaScriptSQLHTML / CSSDjangoFlaskREST APIsBootstrap

AI Agents & LLMs

where the magic gets engineered

LLM IntegrationRAGPrompt EngineeringAgent WorkflowsVector Databasesn8nAnthropic Claude

AI/ML & Data Science

models, metrics & frames

NumPyPandasScikit-learnOpenCVCNNNLPFeature EngineeringModel EvaluationMatplotlibSeaborn

Automation & Scraping

if it's repetitive, it's gone

SeleniumApifySMTP / Email AutomationWeb Scraping

Databases & DevOps

where it all runs

MySQLMongoDBGitGitHubAWSGCPLinux

04/ About

Mumbai · Mostly caffeinated

The human behindthe cron jobs.

I'm Mangesh — an AI Automation Engineer and Data Analyst based in Mumbai, working remote-friendly. My work lives in two places at once: building the systems — LLM agents, RAG pipelines, end-to-end automation — and reading what those systems produce — dashboards, metrics, and decisions people actually act on.

At MCM BPO, I designed and shipped an AI Receptionist that answers inbound calls end-to-end: LLM, speech-to-text and text-to-speech wired into a custom web interface, giving the front desk 24/7 coverage and handing people their time back. At Jolly Clamps, I owned an internal automation portal from whiteboard to deployment, replacing a swamp of spreadsheets with one self-service system.

I hold a B.E. in Computer Science & Engineering with an AI/ML specialization from the University of Mumbai. My operating principle is simple: if a task gets done twice, it should be a script. If it gets done three times, I'm mildly offended it still exists.

0%Manual workflow time cut
0/7Front-desk AI coverage
0+Systems shipped to production
0Spreadsheets I miss

Education

B.E. Computer Science & Engineering

AI/ML Specialization — Theem College of Engineering, University of Mumbai

2021 — 2025

Certifications

  • AWS Academy Graduate — Data Engineering
  • Accenture — Data Analytics & Visualization Job Simulation
  • 100 Days of Code — Complete Python Pro Bootcamp (Udemy)

// Bonus — the portfolio is talkative. Try typing.

mangesh@portfolio — ~/interactive
guest@mangesh:~$

05/ Contact

Usually replies < 24h

Got a repetitive task?Let's kill it.

I'm open to remote roles and freelance automation builds. If it's manual, repetitive, and eating your team's hours — it's exactly my kind of problem.

LocationMumbai, India
StatusOpen to remote
Phone+91 93738 71301
LinkedIn GitHub Email me