$

I build multi-agent AI that survives production.

Senior Data Scientist and AI Systems Architect. I ship production multi-agent systems, MCP servers, and RAG pipelines that real users depend on every day.

Based in New Delhi 3 yrs experience MSc Data Science, London
~/talib/about.py
class Engineer:
  name = "Mohd Talib Akhtar"
  role = "AI Systems Architect"
  building = [
    "multi-agent systems",
    "MCP servers",
    "RAG pipelines",
    "voice agents"
  ]
  status = "shipping daily"
# always learning, always building

The gap between demos and production is where I live.

The best technology is invisible. It just works.

I am a Data Scientist and AI Systems Architect specialising in multi-agent architectures and production-scale GenAI applications. My focus is making powerful, complex AI systems feel simple and natural to use.

I have learned from working across data analytics, program analysis, and production AI that the hard part is not building models. It is building solutions that people actually adopt, trust, and rely on in real environments where things break.

Currently building multi-agent architectures where GenAI moves from demos to production. I am deep in this space because we are still in the early innings of figuring out what these tools can really do, not the hype stuff, but the practical applications that matter.

profile AI Architect
focus Multi-Agent & GenAI
education MSc Data Science
languages EN, HI, UR
location New Delhi, IN
hobbies Chess, Open Source
AI & Agents
Multi-Agent MCP Pydantic AI LangChain RAG
LLM & NLP
OpenAI Anthropic Neo4j Vector DBs
Languages
Python TypeScript SQL PySpark
ML & Data
PyTorch TensorFlow Pandas Databricks
Cloud & Ops
AWS Azure GCP Docker Logfire
Visualisation
Power BI Tableau Streamlit Plotly

Where I have been building.

Feb 2025 — Now
Senior Data Scientist
Zavmo · Bangalore, India

Building production-scale multi-agent AI systems and GenAI applications at an AI-first startup. Architected the full backend from scratch.

  • Architected multi-agent AI system with specialised agents for user guidance, data processing, and workflow automation, serving real users end to end.
  • Engineered MCP server as foundational middleware enabling real-time synchronisation between AI agents, product services, and external APIs.
  • Built end-to-end RAG pipeline using Neo4j graph database and Pydantic AI for structured LLM outputs and context-aware retrieval.
  • Shipped real-time voice pipeline integrating STT and TTS with streaming AI agents for low-latency conversational experiences.
  • Created behavioural analytics dashboard converting user interaction signals into actionable insights for continuous model optimisation.
Feb 2024 — Dec 2024
Data Analyst
Divas World Ltd · Manchester, UK

Led data analytics initiatives for a global e-commerce company, driving business growth through data-driven insights.

  • Analysed customer data from Salesforce CRM using SQL and Python, achieving 15% increase in sales conversion and 20% reduction in churn.
  • Led A/B testing and hypothesis testing to validate product changes, delivering insights to product teams.
  • Built interactive Power BI dashboards translating complex data findings into clear insights for stakeholders.
  • Automated data extraction and transformation using Python and SQL, reducing manual prep time by 40%.
Jul 2022 — Aug 2023
Program Analyst
WNS Global Services · Delhi, India

Built data infrastructure and ML pipelines for enterprise clients, establishing the foundation for advanced analytics.

  • Built and maintained production data pipelines using Databricks, Python, and PySpark for large-scale processing.
  • Developed ML model pipelines, SQL-based warehousing solutions, and feature engineering workflows.
  • Engineered automated data quality validation systems, improving reliability by 10%.

Experiments & open source.

01 / PrepareMe
CASE STUDY

AI Interview Simulation Platform

Multi-agent system running technical, introductory, and cultural fit interview rounds with context-aware feedback. Agents coordinate through task routing, and RAG provides role-specific context. Built solo end-to-end.

Multi-Agent RAG Pydantic AI Python
02 / MCP Framework
WIP

Multi-Agent Orchestration

Scalable multi-agent system with coordinated AI agents collaborating on complex tasks. Implements agent routing, real-time communication, and structured outputs using the Model Context Protocol.

MCP Agents Python
View on GitHub →
03 / Cold Email Gen
OPEN SOURCE

Intelligent Email Generator

GenAI tool using LLMs to generate personalised, context-aware cold emails from a job page link. Uses prompt engineering and few-shot learning for high-quality outreach at scale.

LangChain Streamlit LLMs
View on GitHub →
04 / Autism Detection
RESEARCH

fMRI Classification Model

Processed time-series fMRI data to classify subjects using deep learning. Compared LSTM, CNN, and SVM approaches. LSTM achieved 76% accuracy on the benchmark dataset.

LSTM CNN PyTorch
View on GitHub →
05 / Market Monitoring
PRODUCTION

Financial Data Quality Pipeline

ETL processes and real-time anomaly detection for financial market data. Designed data quality checks and dashboards for monitoring and reporting across multiple data sources.

Azure PySpark Tableau
View on GitHub →
06 / NLP Experiments
OPEN SOURCE

NLP Code Library

Collection of NLP experiments, text classification models, and embedding pipelines. Open source code from my time exploring different ML approaches for text understanding.

NLP Transformers Python
View on GitHub →

Got an AI problem that feels unsolvable?

I love talking about multi-agent systems, production AI, and how to take cool demos and make them ship. Reach out.

// LINKEDIN
/in/mohdtalibakhtar
// GITHUB
@mohdtalibakhtar
// LOCATION
New Delhi, India