Ml Engineer Machine Learning
Resume:
Suraj A Bennur
AI/ML Engineer Python
AI/ML engineer with 12+ years of experience delivering production-grade solutions in computer vision, natural language processing, generative AI, and real-time analytics. Expertise in developing machine learning models, deep learning architectures, and end-to-end pipelines using TensorFlow, PyTorch, and Scikit-learn. Strong background in Python development, API integration (Flask, FastAPI), and backend systems for scalable deployments. Skilled in cloud platforms (AWS, GCP, Azure) for training, deployment, and monitoring of AI models. Recognized as ‘Engineer of the Quarter’ at Microsoft for significant contributions to team projects and improved intent detection accuracy by 35% through multi-agent systems. Hard Skills
● Programming Languages: Python, SQL
● Frameworks & Libraries: Django, Flask, FastAPI, Pandas, NumPy
● ML Frameworks: PyTorch, Tensorflow, Hugging Face, OpenCV, Librosa
● Natural Language Processing: RAG, Langchain, LangGraph, RNN, LSTM, LLM, BERT, GenAI, Chatbot, Speech-to-Text, Text-to-Speech
● Databases: MySQL, PostgreSQL, MongoDB
● DevOps & Cloud: Git, GitHub, Docker, Jenkins, AWS, GCP
● Testing: PyTest, UnitTest
● Security & Authentication: OAuth 2.0, OpenID Connect, SSO Soft Skills
● Mentoring
● Team Collaboration
● Client Interaction
● Problem Identification and Resolution
***********@*****.***
https://www.linkedin.com/in/suraj-bennur
905 W President George Bush Hwy Apt 12312, Richardson, TX 75080 WORK EXPERIENCE
Fujitsu Network Communications Dallas, TX
AI Engineer, Team Leader 03.2024 - Present
● Designed an end-to-end RAG-based knowledge retrieval system, revolutionizing disorganized documentation search with accurate retrieval, decreasing average resolution time by 3 hours and improving the search experience for stakeholders.
● Architected and executed end-to-end AI/ML pipelines across computer vision, diffusion-based AI workflows, and low-latency analytics, scaling retail and personalization applications to support 100K+ users with a 30% reduction in processing time.
● Orchestrated computer vision pipelines utilizing YOLOv8 and Vision Transformers
(ViT) for real-time object tracking, resulting in a decrease of 20 manual labor hours each week.
AI Engineer, Lead Developer 10.2020 – 03.2024
● Devised and orchestrated a multi-agent architecture leveraging LLMs (GPT-4o, GPT-4o-mini, Cohere), achieving a 35% improvement in intent detection accuracy and 40% reduction in misclassification rates.
● Pioneered a virtual try-on feature utilizing OpenPose and ControlNet, directing a team of 3 engineers to achieve 98% success in the first month of product launch.
● Constructed and fine-tuned intelligent classification models supporting recommendation systems, wardrobe organization, and event detection. Senior AI Engineer 01.2018 – 09.2020
● Enhanced model throughput by 40% through optimizing inference pipelines on GPU infrastructure and cloud-based computer platforms, which improved model performance.
● Collaborated with cross-functional teams to conceptualize user-centric AI features, increasing user engagement by 25% and generating 15+ actionable insights adopted into product roadmaps.
Technologies: Python, PyTorch, TensorFlow, Langchain, LLM, RAG, Vectors, BERT, Cross- Encoding, YOLOv8, ViT, Diffusion Models, OpenCV, Mediapipe, NumPy, Pandas, FastAPI, Flask, REST APIs, Docker, Runpod (GPU compute), NVIDIA GPU, GPT-4o, GPT-4o-mini Ericsson Santa Clara, CA
AI Engineer 06.2016 – 08.2016
● Developed and deployed a single sign-on (SSO) framework using OAuth 2.0 and OpenID Connect, enabling over 5,000 Ericsson employees to securely and seamlessly access MediaFirst Cloud TV services through Ericsson Identity.
● Guided the integration of natural language chatbots and voice synthesis systems to deliver personalized conversational AI solutions, improving retrieval accuracy by 80% and expanding multilingual support to 2 languages.
● Created an agentic, tool-calling anomaly detection system using knowledge graphs, machine learning, and LLM reasoning to correlate real-time device logs, metrics, and documentation accelerating triage and sharply reducing manual investigation overhead.
● Engineered an interactive chatbot that improved suggestion precision by 73% and decreased the average chat session length by 20%, enhancing user satisfaction and overall chatbot efficiency.
● Integrated vector search (Pinecone) with semantic search, query rewriting, and metadata filtering, boosting multilingual recommendation relevance by 80% across English and Swedish.
● Crafted and enhanced an AI-driven voice cloning platform with 15 uniquely modeled voice agents for natural speech synthesis.
● Refined backend pipelines with FastAPI and Python, reducing response latency by 45% and enabling real-time text-to-speech with 99.9% uptime.
● Applied advanced audio preprocessing (Librosa, PyTorch) and fine-tuning with 80 minutes of speaker data per voice, improving speech synthesis accuracy and naturalness.
Technologies: Cohere, Pinecone, LangChain, FastAPI, Python, PyTorch, So-VITS.SVC, Hugging Face TTS, NumPy, Librosa, Git, REST APIs, Webhooks Parasoft Bengaluru
AI Developer 08.2012 - 06.2015
● Engineered a real-time dice recognition system utilizing live camera feeds, achieving 96% outcome detection accuracy and reducing error rates in validation scenarios by 30%.
● Assembled and deployed a vision-based pipeline with YOLO and OpenCV, improving dice face recognition precision by 23% and enabling sub-second inference speeds.
● Constructed Python scripts leveraging NumPy to automatically compute dice totals, decreasing calculation errors by 65% due to robust error handling and increased game accuracy.
● Streamlined inference and image-processing workflows in Python and NumPy, reducing model latency by 45% and increasing throughput to handle 1,500+ real-time video frames per second.
● Fortified ParaSoft C/C++ software with 85+ static code analysis rules guided by CERT standards, boosting application security for customers and curbing vulnerabilities by 18%.
Technologies: YOLO (Object Detection, Oriented Bounding Boxes), Python, OpenCV, NumPy, Real-Time Video Processing, Error Handling Logic Education
California State University Fullerton, CA
Master’s Degree Computer Science 07.2015 – 12.2017 Visvesvaraya Technological University Karnataka, India Bachelor’s Degree, Instrumentation Technology
Contact this candidate