Skip to content
Delta AI Engineering

About · Founder

REI I. ZAIDI

AI engineerApplied data scientistSoftware architectResearcher

Founder of Delta AI Engineering and Phoenix Group · Karachi, Pakistan

REI I. ZAIDI designs enterprise AI systems that connect models, knowledge, tools and business workflows, with reliable execution, human oversight and measurable outcomes.

Her career began in December 2010 in .NET delivery teams and moved through enterprise SharePoint, Azure and Power Platform architecture to production generative AI. At Infixio she is a Senior AI Engineer and Applied Data Scientist, leading multi-agent LLM, retrieval and computer-vision platforms across Claude, GPT, Gemini and Mistral, with evaluation gates, tracing and cost guardrails.

REI I. ZAIDI, founder of Delta AI Engineering

Why “Delta”

In engineering, Δ means change. We exist to produce a specific, measurable change in how an organisation works, and to prove it happened.

“My mission is to translate research into accountable AI systems, sustainable businesses and international collaborations.”
REI I. ZAIDI, founder
  • PEOPLE

    Systems are built for the people who use them and the people they affect. Humans approve what matters.

  • TECHNOLOGY

    Proven tools, typed contracts and explicit state, chosen for the problem rather than the trend.

  • SOLUTIONS

    Working software in production, not slideware. One complete slice before broad scope.

  • IMPACT

    A baseline at the start, evidence at the end, and monitoring that keeps the result honest.

In numbers

The work behind the name

Years in software
11+
Including 5+ years in AI
Systems catalogued
27
From live products to concepts, each with a status
Screening records analysed
4,460
Autism-screening BI, Phoenix Minds, 2024
Languages in product
58
DRIS and LUMINA, with right-to-left support
Public Mars datasets
21
NASA and ESA imagery, MARS-MINDS v2
Manuscripts in preparation
6
Elsevier Q1 and IEEE targets

Expertise

What is proven, what is built, what is next

Skills are grouped by where they have been used, so production experience is never confused with study.

In production

Delivered in production or client work

  • LangGraph multi-agent systems
  • Human-in-the-loop control planes
  • MCP servers and tool use
  • Enterprise RAG with semantic reranking
  • Provider-agnostic model routing
  • Regression-evaluation release gates
  • FastAPI, PostgreSQL and pgvector
  • Docker, Kubernetes and GitHub Actions
  • LangSmith observability
  • .NET, SharePoint, Azure and Power Platform

Products and research

Built in products, prototypes, research or coursework

  • Refusal-aware routing with human review
  • BM25 + FAISS hybrid retrieval
  • Neo4j knowledge graphs
  • PEFT, LoRA, QLoRA and DPO
  • Computer vision for planetary imagery
  • NVIDIA Triton serving
  • A2A protocol hand-offs
  • Multilingual and right-to-left NLP

Exploring now

Current study

  • Deep Agents
  • Temporal durable execution
  • gVisor sandboxing
  • Inspect AI evaluations
  • OpenTelemetry tracing
  • vLLM, batching and quantisation

Career

5+ years in AI, built on 11+ years in software

Applied AI in production is the focus. Enterprise software engineering, practised since December 2010, is the foundation underneath it.

  1. 2007 – 2010 · Education

    BS Computer Science

    Hamdard University, Karachi

    The foundation for an engineering career that began the month she graduated.

  2. Dec 2010 – Jul 2015 · Engineering

    Junior Software Engineer → Software Engineer

    Techerz IT Consultancy Firm

    ASP.NET and C# delivery, testing and deployment; later led .NET web teams, wrote specifications, ran technical interviews and set coding standards.

  3. Aug 2015 – Jan 2019 · Engineering

    Senior Software Engineer

    ExaTex Enterprise Solutions

    Full-lifecycle delivery of SharePoint and Project Server solutions, business-process automation, intranets and portals.

  4. Feb 2019 – Oct 2023 · Engineering

    Lead Software Engineer → Software Architect

    Learning Pitch

    Led enterprise teams, then architected secure, scalable systems on Azure, SharePoint and the Power Platform, with pre-sales and technical consulting.

  5. Jan 2022 – present · Venture

    Founder

    Phoenix Group

    A research and AI group spanning applied research, shared AI infrastructure, healthcare and special-education AI, and commercial AI products.

  6. Jan 2023 – Dec 2025 · Education

    MSc Artificial Intelligence

    NED University of Engineering and Technology

    Department of Computer and Information Systems Engineering. Thesis: MARS-MINDS, deep learning analysis of spaceborne sensor data for the origins of Martian dust storms.

  7. Nov 2023 – present · AI practice

    Senior AI Engineer and Applied Data Scientist

    Infixio

    Architecture and hands-on delivery of multi-agent LLM, RAG and computer-vision platforms, with structured-output validation, regression evaluation and cost controls.

Ventures

Phoenix Group

Founded in January 2022, Phoenix Group brings together the founder's research, shared AI infrastructure, healthcare and education products. Every product carries its real status.

Research and development

Phoenix Research and AI Systems

Applied AI research and technology evaluation, home of the MARS-MINDS programme.

  • MARS-MINDS v2 · In development

Shared AI infrastructure

Phoenix Nexus

Model orchestration, evaluation tooling and platform services shared across the group.

Healthcare and special-education AI

Phoenix Minds

Clinical and education tools built so that professionals, not models, make the decisions.

  • DRIS · v1.0 live
  • Zeta-9 · Operating network
  • RAHAT-AI · Prototype

phoenixmindsai.com (opens in a new tab)

Commercial AI products

Phoenix AI-AAS

AI-as-a-service products for education, finance and enterprise operations.

  • TALEEM · MVP delivered
  • LUMINA · Active build
  • Phoenix Sentinel · Prototype delivered
  • Phoenix Horizon · Active build

Research · MSc thesis

MARS-MINDS

Martian Intelligent Navigation and Decision System

Deep learning on orbital imagery to classify terrain, find habitat sites, judge landing hazards, plan missions and forecast dust storms. Version 2, in development since March 2026, coordinates five specialist models with a reinforcement-learning controller across 21 public NASA and ESA datasets.

25-class mission-planning classification
99.91%
EfficientNetV2 + ConvNeXtV2
Safe-landing hazard identification
F1 0.97
MarsLanderNet, DenseNet121
8-class terrain and dust-storm classification
94.00%
InceptionV3
Habitat-site detection
mAP 0.681
AstroVisionNet, Faster R-CNN

Results as reported by the author from thesis experiments. NASA, JPL and ESA imagery is public data; its use does not imply affiliation with any space agency.

Research pipeline

Six manuscripts in preparation

Targeting Elsevier Q1 (Scopus) and IEEE venues. None is presented as published.

  1. 01 · In preparation

    Automated Mars Feature Classification Using InceptionV3 for Enhanced Planetary Surface Analysis

    InceptionV3 · 8-class terrain

  2. 02 · In preparation

    MarsLanderNet: A DenseNet-Based Framework for Autonomous Safe Landing on Martian Terrain

    DenseNet121 · landing hazards

  3. 03 · In preparation

    AstroVisionNet: AI-Driven Martian Terrain Classification for Habitat Construction and Dust Storm Risk Mitigation

    Faster R-CNN + ResNet-50 · habitats

  4. 04 · In preparation

    Parallel Dual-Backbone Image Classification for Mars Mission Planning: EfficientNetV2 & ConvNeXtV2

    Dual-backbone fusion · 25-class

  5. 05 · In preparation

    Mars Dust Storm Forecast

    Faster R-CNN + ConvLSTM · multi-sol forecasting

  6. 06 · In preparation

    MARS-MINDS v1 and v2: integrated five-model frameworks

    Navigation, habitat, landing, planning, dust risk

Earlier certificates and publications appear under the name Ruksh E Ibadat.

Recognition and credentials

The paper trail

Recognition and roles

  • ASPIRE Pakistan, National Idea Bank IV. Pre-incubation Rank #1 finalist for Zeta-9, among 24 start-ups selected from 800 applications.
  • AQLENTIS AI residency. Co-Team Lead of a 12-week AI residency with NCBC and HPCC, NED University (Batch 1, from September 2026).

Degrees and training

  • MSc Artificial Intelligence, NED University of Engineering and Technology (2023–2025)
  • BS Computer Science, Hamdard University (2007–2010)
  • GenAI to Agentic AI: Full-Stack Development · NCBC, HPCC and NED University (2026)
  • Postgraduate Certificate in Data Analytics and Data Mining · NED Academy (2022)
  • Diploma in Early Childhood Education and Development · IECED Karachi (2017–2018)
  • Skills-based CBT and DBT courses · MedicPro International Academy (2025–2026)

Training certificates inform the design of healthcare and special-education products. They do not confer clinical licensure.

Everyday stack

Tools she reaches for

  • Python
  • TypeScript
  • LangGraph
  • MCP
  • FastAPI
  • PostgreSQL
  • pgvector
  • Neo4j
  • PyTorch
  • Docker
  • Kubernetes
  • GitHub Actions
  • LangSmith
  • Azure
  • Next.js
“Neither a jack of all nor a king of one, but a king of all.”
REI ZAIDI, personal motto

Where we work

Based in Karachi, Pakistan. Working across time zones.

We work remotely with teams in other regions, with overlap hours agreed at the start of every engagement.

Tell us the change you need. We will tell you how we would measure it.

A short first conversation, a written summary afterwards, no obligation.