➜ ~ whoami
Mohammed Uzair
Full-cycle ML engineer building GenAI, voice and computer-vision systems across text, image and video — from data strategy and model design to production deployment.
~3
years building AI in production
100
RTSP camera streams handled (airport CV)
8.64
B.Tech CGPA, Osmania University
GATE '24
qualified
01 about.py
from humans import Engineer class Uzair(Engineer): name = "Mohammed Uzair" role = "Senior Gen AI / Computer Vision Engineer" based_in = "Hyderabad, Telangana, India" education = "B.Tech CS, Osmania University (2024)" focus = ["Agentic AI", "RAG", "Voice agents", "Computer vision"] published = "ICDSAI 2023 research paper" def mission(self): return "Apply trending research to build real-world, impactful AI systems."
02 skills.json
"languages"
"genai_agents"
"vector_stores"
"ml_cv"
"serving_backend"
"devops_tools"
"data_cloud"
03 experience.log $ git log --oneline
Senior Gen AI Engineer (SE-3) @ Wells Fargo
Conversational Voice Bot GenAI · Voice Agents · LLM · STT · TTS
- Built a skills-based multi-agentic system with Google ADK for conversational querying on Google BigQuery.
- Designed multiple conversational voice agents in Google CX Studio, integrating LLMs, STT and TTS for real-time voice interaction.
- Built evaluation frameworks (Evals) in CX Studio and ADK to assess response quality and flows and find areas to improve.
- Led weekly production releases independently: deployment, testing, debugging and release validation.
- Developed a RAG-based intent classifier that routes queries to the right agent/skill, improving routing accuracy and cutting unnecessary agent execution.
AI & Computer Vision Engineer (SDE-1) @ Master Works Software
Multimodal GenAI Platform GenAI · RAG · LLM · LangGraph · Nov 2024 – Mar 2026
- Built a multi-RAG chatbot over documents, PDFs, PPTs, spreadsheets, images, URLs and databases, using OCR, Fitz and VLM pipelines for extraction.
- Implemented Natural Language to SQL with a Chat-with-Database module; spreadsheet queries via SQL, Pandas and unstructured querying.
- Shipped production conversational AI with memory, tool use, guardrails and real-time streaming; added MCP, multi-model support, infinite history and long-term memory.
- Designed Self-Adaptive RAG (up to 50% better results) and a custom retriever with reranking and multi-query retrieval.
- Used FAISS and ChromaDB for vector search, improving search efficiency by 40%.
- Used LangGraph / LangGraph Studio across modules for 70% performance gains, with custom multi-agent architectures.
- Deployed models with TGI, vLLM, Ollama, Hugging Face and Unsloth: 30% faster inference, 50% lower latency.
- Built end-to-end real-time Voice AI pipelines (STT + TTS); deployed across on-prem and Oracle Cloud.
- FastAPI endpoints, JWT-based authorization, Docker/Poetry/UV deployment, Prometheus + Grafana monitoring.
Custom Face Recognition System for Airport CV · Self-registration · Aug – Oct 2024
- Self-registration face recognition on NVIDIA DeepStream with GST Nvtracker, custom YOLO, RetinaFace and ArcFace.
- Fine-tuned detectors with NVIDIA TAO Toolkit, improving accuracy by 25%.
- Qdrant for vector similarity search, Kafka for event streaming, Docker for packaging.
- Handled around 100 RTSP cameras in real time.
Face Recognition & Analytics System for Gym CV · Real-time analytics · Mar – Aug 2024
- Real-time recognition with entry/exit tracking, employee presence, emotion analysis and heatmaps on DeepStream.
- Kafka streaming across networks, Dockerized pipeline, ChromaDB face search cutting retrieval latency by 30%.
- Worked with RTSP streams from 32 cameras using PeopleNet, RetinaFace, ArcFace and an emotion model.
04 projects/
def ai_interviewer()
Takes a job description and conducts the interview. Whisper Live (STT), Qwen (LLM) and Coqui TTS, with an emotion metric and scoring for each Q&A pair.
def face_emotion_analysis()
Face detection, recognition and emotion analysis on images and video: YoloFace for detection, ArcFace for recognition, facial behaviour analysis for emotion.
def plant_disease_detector()
CNN model behind a Flask web app. Upload a leaf image, get the disease prediction, related info and suggested supplements.
def disaster_early_warning()
CNN disaster detector that watches live camera video and emails response teams automatically on detection.
def aspect_sentiment()
Extracts product reviews and comments and outputs aspect-level sentiment scores using a custom aspect extractor and a trained sentiment model.
05 certs.md & research
# Certifications
- Microsoft Certified: Azure AI Fundamentals
- Python Certified Associate Programmer — Cisco
- Python Fundamentals & Programming — NASSCOM / MeitY
- HackerRank: Python, SQL, Problem Solving (Basic)
# Research & Awards
- “System Sound Control using Gesticulations” — sole author, presented at ICDSAI 2023, to be published in Springer’s journal
- Best Team Player 2024–25 (GenAI product)
- Qualified GATE 2024
# Education
- B.Tech CS, Osmania University — CGPA 8.64 (2020–24)
- Intermediate, MS Junior College — CGPA 8.47 (2018–20)
- Our Own English High School, Sharjah — A+ throughout
06 contact.sh
$ echo $EMAIL
$ echo $LINKEDIN
$ echo $LOCATION
Hyderabad, Telangana, India
$ # open to AI Engineering & Agentic AI roles