RASIKH natiq

RASIKH natiq

Full Stack AI Engineer specializing in RAG systems and Multi-Agent workflows.

Based in: Islamabad, PakistanMost recently: Full Stack AI Engineer, Stixor TechnologiesExperience: 4 years

Hire me

Experienced AI Engineer with a strong background in building real-time voice agents and scalable Docker-based AI deployments. Currently pursuing a Master's in AI, he focuses on delivering context-aware autonomous solutions from research to production.

Experience

Full Stack AI Engineer

Stixor TechnologiesIslamabadworkApr 2025 – Sep 2026

Summary

Owned AI products end to end as a full stack AI engineer, covering backend services, AI logic, system architecture, and deployment. Designed and implemented Retrieval-Augmented Generation (RAG) systems that grounded LLM responses in relevant data and improved contextual accuracy. Built and deployed multi-agent systems for complex task coordination and autonomous decision-making. Developed and deployed real-time voice AI agents capable of natural conversation, task execution, and integration with external services for automated workflows. Architected backend and AI systems, built the APIs connecting them to client apps and third-party services, and guided the team on scalable AI design. Containerized AI applications with Docker to deliver scalable, consistent, environment-independent deployments.

What I did

  • Converted the entire calling agent system to an asynchronous architecture.
  • Improved the latency of individual tools within the agent system.
  • Implemented background sounds, such as typing, to maintain user engagement during processing.
  • Implemented a multi-agent system for the Malakah project to solve user problems using domain-specialized legal agents.
  • Implemented a mechanism to analyze user behavior changes by extracting a JSON of the answer and sentiment from the LLM with each response.
  • Implemented runtime validation functions to verify customer emails before booking meetings.
  • Implemented a retry loop for email verification that uses regex, LLM extraction from transcriptions, and user confirmation.

Results

  • Optimized a calling agent's response time from 7 seconds down to 1.5 seconds.
  • Improved RAG retrieval for MARI Energies by writing prompts that enabled agents to retrieve relevant information from databases via SQL queries.
  • Reduced calling agent latency from 7s to 1.5s
RAGLLMMulti-agent systemsVoice AIRESTful APIsDockerBackend ArchitectureAsynchronous ProgrammingSQLPrompt EngineeringPydanticFew-shot promptingSchema Validation

AI Developer

DigiCare IT ConsultantsGujranwalaworkSep 2023 – Mar 2024

Summary

Developed AI algorithms to analyze and predict trends in financial time-series data. Applied statistical and machine learning models for accurate forecasts, and used data visualization to communicate insights. Integrated AI models and analytics into the company's ERP system, bringing predictive insights directly into business workflows. Managed end-to-end project workflows and evaluated and optimized model performance to improve accuracy.

What I did

  • Integrated financial trend forecasts into a custom ERP developed specifically for factories.

Results

  • Developed an algorithm to predict total sales, costs, and expenses for upcoming months using Liquid Neural Networks.
  • Achieved 80% accuracy using Liquid Neural Networks for financial forecasting.
  • Improved prediction accuracy from 70% to 80% by transitioning from LSTM to Liquid Neural Networks.
Machine LearningFinancial Time-series DataStatistical ModelsData VisualizationERP IntegrationLiquid Neural NetworksLSTMTime-series Forecasting

AI Researcher and Developer

BZ TechnologiesNorthamptonshire, UKwork2021 – Mar 2023

Summary

Led the development and optimization of AI models across Machine Learning, Computer Vision, and Natural Language Processing. Conducted in-depth experimentation and evaluation across research domains to improve model accuracy, robustness, and generalization. Collaborated closely with cross-functional teams to address data quality issues and ensure high-performance model deployment. Designed and implemented custom pre-processing pipelines tailored to project-specific data characteristics.

Results

  • Improved breast cancer detection accuracy from 85% to 96% by testing multiple ResNet architectures and creating data variations.
Machine LearningComputer VisionNatural Language ProcessingData Pre-processingResNet

Skills

Technical

PythonPython
GitHubGitHub
LLMs
Visual Studio Code
RAG
API Development
RESTful APIs
Multi-Agent Systems
Full Stack Development
Voice AI Agents
Computer Vision
NLP
DockerDocker
Visual Studio
C++C++
Feature Engineering
Windows Desktop Applications
SQLSQL
Liquid Neural Networks
LSTM
ResNet
Pydantic
Schema Validation
Asynchronous Programming
Ollama

Areas of expertise

Cross-Functional Collaboration
Team Leadership
Project Management

Projects

MARI Energies RAG System

personal
Developed a RAG system that utilized agents to retrieve relevant information from databases by generating SQL queries through specific prompting.
SQLRAGAgents

Malakah Project

Developerpersonal
Developed a multi-agent system where specialized agents focused on specific law domains and a manager agent coordinated tasks to provide users with all possible scenarios.

Local Sentiment Analysis Experiment

personal
Ran a BERT-based model locally to accumulate user sentiment after calls, avoiding additional LLM API calls.
OllamaBERT

Education

Master's in Artificial Intelligence

FAST University of PakistanArtificial Intelligence2026

Bachelor's in Artificial Intelligence

Air University of PakistanArtificial Intelligence2024

Recognition

Certifications

Prompt Engineering

IBM2024

AI Applications with Python and Flask

IBM2024

Python for Data Science, AI and Development

IBM2024

Microsoft Certified: Azure AI Fundamentals

Microsoft2022

Generative AI: GANs to CLIP

DeepLearning.AI2022

Generative AI with LLMs

DeepLearning.AI2022
Hire RASIKHReach out about a role, a contract or a conversation.For recruiters

RASIKH's twin is AI, it can make mistakes.

More people on Kanvis

Browse more peopleCreate your own Kanvis