Hi, I'm Vaibhav. I build AI that ships
and scales in production.

Machine Learning Engineer & Generative AI Specialist with 9+ years designing, deploying and scaling AI systems. Generative AI Expert at Deutsche Telekom by day — building MagentaAI, FlexiBot and askT. By night, the founder of CVCL.online — AI-tailored CVs and cover letters for honest job applications.

See my work LinkedIn
Years building AI
0+
Production-grade PoCs
0+
Users served
0K
Cloud platforms
0
01

About

Generalist depth across the AI stack — from data to deployment to monitoring.

I'm an ML engineer who's spent the last decade quietly turning research and prototypes into systems that real users rely on. My happy place is the intersection of Generative AI, backend architecture and LLMOps — where the interesting problems live and where cleverness alone isn't enough.

At Deutsche Telekom I architect and ship AI products serving the company's workforce and customers. I've built chatbots that handle 80k users, multimodal real-time agents, RAG pipelines feeding vector DBs, LLM load balancers, evaluation frameworks and multi-agent orchestration — all in compliant, enterprise-grade environments (GDPR, EU regulations, workers' councils).

Before that I built credit-risk models, churn predictors and recommendation engines as a Data Scientist; and before that, I started out on time-series trading systems and full-stack apps. That breadth shows up in how I design things: I optimise for what survives contact with production.

What I do best

  • Design and ship production GenAI systems — RAG, agents, fine-tuning, evaluation
  • Stand up LLMOps end-to-end: routing, caching, monitoring, cost & latency control
  • Architect scalable Python microservices with FastAPI, Docker, Kubernetes
  • Build data pipelines into vector DBs (Postgres + pgvector, Pinecone, etc.)
  • Drive compliance-aware AI — GDPR, EU AI Act, enterprise governance
  • Translate ambiguous business asks into clear, measurable AI roadmaps

Currently

Architect & GenAI Expert at Deutsche Telekom — working across NatCos and vendor teams to roll out MagentaAI features at scale. Outside of that, founder & sole operator of CVCL.online, an AI-tailored CV and cover-letter platform with 1,200+ CVs generated.

02

Experience

A decade of building, shipping and operating AI in the wild.

  1. Aug 2022 — Present Remote · Germany / India

    Generative AI Expert & Architect · Deutsche Telekom

    Leading GenAI initiatives across five flagship projects — MagentaAI, Multimodal Magic, FlexiBot, askT and GK HUB — owning architecture, build and rollout end-to-end.

    • Designed and shipped chat systems serving 80,000+ users, with multi-agent orchestration, LLM load balancing and evaluation frameworks.
    • Built real-time multimodal agents that talk to users over audio, powered by live RAG.
    • Fine-tuned GenAI, agentic and BERT models; ran automated data annotation pipelines using LLMs.
    • Stood up the LLMOps stack — proxies, observability, Prometheus/Grafana dashboards — to keep response quality and cost in check.
    • Drove GDPR, EU and workers' council compliance across every AI rollout.
  2. Apr 2019 — Apr 2022 Pune, India

    Data Scientist · Tyraa Systems

    Built ML systems across credit risk, churn and recommendations — owning the full lifecycle from data to deployed APIs.

    • Created credit-risk models to identify optimal parameters for risk evaluation, deployed via APIs on VMs.
    • Trained churn prediction models that fed directly into business decision-making.
    • Shipped a domain-specific recommendation engine — drove a 40% lift in session time via A/B tested strategies.
  3. Sep 2017 — Mar 2019 Hyderabad, India

    Machine Learning Engineer · Konnectsu Technologies

    Built the KYC modelling stack for a financial institute, plus NLP pipelines for customer insight.

    • Trained and deployed ML models that improved decision-making accuracy by 20%+.
    • Built end-to-end ML workflows — collection, preprocessing, training, cloud deployment on Digital Ocean.
    • Shipped NLP pipelines for text classification and sentiment analysis.
  4. Jun 2016 — Sep 2017 Pune, India

    Junior Engineer · Affingo Technologies

    Where it all started — time series for finance, and full-stack mobile/backend on Python + Java.

    • Saralstocks.com — built a univariate time-series engine on BarChart & Yahoo Finance data for stock-trading strategies.
    • Khojpal — Android app in Java with a Python/WebPy backend connecting local businesses to users.
03

Selected projects

A few of the things I've built recently that I'm proud of.

Founder · Live product

CVCL.online

AI-tailored CVs + cover letters for honest job applications. Upload your CV once, paste any JD, get a tailored CV and cover letter in PDF and editable DOCX in under 60 seconds — reframed in the role's vocabulary, never inventing experience. Also includes a real-time AI voice interviewer, a free ATS-readability check, and parallel job-board scanning across 30+ company boards (Greenhouse, Ashby, Lever). 1,200+ CVs generated, 4.8 avg rating.

  • FastAPI
  • LLM tailoring
  • PDF + DOCX
  • Voice interviewer
  • ATS scoring
  • Razorpay
  • DigitalOcean (Frankfurt)
Deutsche Telekom

MagentaAI

Fine-tuning GenAI, Agentic and BERT models for Telekom's flagship AI initiative. Automated data annotation pipelines, vendor collaboration across NatCos, and GDPR/EU-compliant rollouts.

  • LLM fine-tuning
  • Agentic AI
  • BERT
  • GDPR
Deutsche Telekom

Multimodal Magic

A real-time multimodal framework that converses with users over audio. Integrated multimodal live APIs with a RAG backend and built the demo flows that took it from idea to live deployment.

  • Multimodal
  • Realtime audio
  • RAG
  • Live API
Deutsche Telekom

FlexiBot

Architected an enterprise-grade chatbot from the ground up — frontend, backend, AI engine, LLM load balancer, evaluation framework and multi-agent runtime — handling intensive real-time load.

  • Multi-agent
  • LLM routing
  • RAG
  • Eval
  • Vector DB
Deutsche Telekom

askT

Scaled AI-powered chat to 80,000 users. Managed OpenAI integrations via proxies for cost & quality metrics, ingested vendor data sources, and ran continuous performance monitoring.

  • 80K users
  • OpenAI
  • Observability
  • Cross-team
Deutsche Telekom

GK HUB

Led deployment of AI components for 10,000 users. Benchmarked multiple vector DBs and LLMs, then shipped scalable Python services with LLMOps techniques to keep latency and cost in check.

  • LLMOps
  • Vector DB benchmarks
  • Python APIs
Tyraa Systems

Recommendation Engine

Built a content recommendation engine on domain-specific text data. Designed A/B testing workflows that proved out a 40% lift in session time on targeted user cohorts.

  • Recommenders
  • A/B testing
  • NLP
04

Skills & toolbox

The tools I reach for, grouped by what I'm building.

AI & ML

  • Generative AI
  • Agentic AI
  • LLM fine-tuning
  • RAG
  • BERT / Transformers
  • Deep Learning
  • Conventional ML
  • NLP
  • AI System Design
  • Evaluation frameworks

Languages & Frameworks

  • Python
  • Shell
  • PyTorch
  • scikit-learn
  • FastAPI
  • LangChain
  • K6

Data & VectorDBs

  • Vector databases
  • pgvector
  • Data engineering
  • Web scraping
  • Data annotation (LLM-assisted)
  • A/B testing

Cloud & DevOps

  • Docker
  • Kubernetes
  • CI/CD
  • AWS
  • GCP
  • Azure
  • Digital Ocean
  • Prometheus
  • Grafana

Practice & ways of working

  • System architecture
  • API design
  • Performance optimisation
  • Security & compliance
  • Technical documentation
  • Stakeholder communication
  • Cross-functional leadership

Languages I speak

  • English
  • Deutsch
  • हिन्दी
  • मराठी
05

Let's build something.

I'm open to senior AI/ML roles, consulting and advisory work — particularly anything that involves taking GenAI from prototype to scaled production. The fastest way to reach me is email.

Email Vaibhav.kadam5757@gmail.com
LinkedIn /in/vaibhav-data-scientist
Phone +91 940 424 1842
GitHub @lurndatascience
YouTube @vaibhav.kadam.data.science
Instagram @trainset_ai
Stack Overflow Vaibhav Kadam
My product cvcl.online
Start a conversation