Scaling Laws in AI: Why Bigger Models Work in Research but Break in Production
artificial-intelligence machine-learning data-science software-development deep-learning
9/6/2025
Read More →
Yuval Mehta
AI/ML Engineer | Deep Learning Practitioner | MLOps Enthusiast
I specialize in building scalable machine learning pipelines, computer vision models, and NLP systems that deliver real-world impact. Currently working on intelligent systems at the intersection of AI research and production-ready deployment.
About Me
Hi, I'm Yuval Mehta 👋
I'm a passionate AI/ML Engineer driven by curiosity and impact. I specialize in developing data-driven solutions that blend machine learning, deep learning, and scalable systems to tackle real-world challenges. With hands-on experience in computer vision, natural language processing, and MLOps, I bring innovative AI models from research to deployment.
🚀 What I Do
- ML System Design – Building and scaling end-to-end ML pipelines using MLOps best practices
- AI Development – Creating production-ready models in computer vision and NLP domains
- Experimentation & Optimization – Fine-tuning models for peak performance and reliability
- Product Impact – Developing intelligent systems that deliver measurable business outcomes
🏆 Key Achievements
- Ranked Top 1% (274th/74,824) in Amazon ML Challenge 2024 with a team of four
- Published research paper on Early Diabetes Prediction at IEEE APCIT 2024
- Improved video processing throughput by 30% at IIT Kharagpur using GNNs and autoencoders
📚 Education
- B.Tech in Computer Engineering
- NMIMS - Mukesh Patel School of Technology, Management and Engineering
- Specialization in Artificial Intelligence
- Focused on: Machine Learning, Deep Learning, Computer Vision, Natural Language Processing, Reinforcement Learning and MLOps
- Core: Data Structures and Algorithms, Computer Architecture and Organization, Operating Systems, Database Management Systems, Computer Networks, Theory of Computation
💻 Technical Skills
Languages:
Python, JavaScript, SQL, C, C++, Java
ML/DL Frameworks:
PyTorch, TensorFlow, Keras, Scikit-learn, XGBoost, OpenCV, Transformers, NLTK, spaCy, LightGBM, CatBoost
AI Agents:
LangChain, Langgraph, CrewAI, A2A, MCP, LLamaIndex
MLOps & Cloud:
MLflow, Docker, Weights & Biases, DVC, CI/CD, ONNX, AWS, GCP, Azure
Web Technologies:
Django, FastAPI, Flask, Streamlit, Node.js, Express.js, HTML, CSS, Streamlit
Databases:
MySQL, MongoDB, PostgreSQL, Supabase, Vector Databases, Redis, Pinecone, ChromaDB, FAISS, SQLite
Automations:
N8N, Make, LangFlow, Zapier, SuperAGI
Big Data & Distributed Computing:
Pyspark, Apache Spark
Specialties:
Computer Vision, NLP, Generative AI, ML Pipelines, Reinforcement Learning, LLM Fine-tuning, AI Automation, Agent-based Systems
Experience
AI/ML Engineer
xLM - Continuous Intelligence - Remote
June 2025 - Present
- Engineered AI agents within cIV (compliance intelligence platform), automating GxP workflows and expanding task coverage by 65%.
- Built LangGraph-based multi-agent systems with retry, memory, and control flows—cutting execution time by 30% and boosting success rates by 40%.
- Embedded traceable logic and collaborated with QA to enhance audit-readiness and system transparency.
AI/ML Intern
xLM - Continuous Intelligence - Mumbai, Maharashtra
January 2025 - May 2025
- Constructed an automated traceability matrix generator—reducing manual effort by 60% and increasing consistency by 45%.
- Deployed real-time QA pipelines and logging systems, enabling 100% traceability in model operations.
Machine Learning Research Intern
IIT Kharagpur - Remote
June 2025 - May 2025
- Conducted experiments with auto-encoders and GNN, boosting video processing by 30%.
- Optimized hyper-parameters, resulting in a 20% increase in model accuracy and efficiency.
Machine Learning / Data Science Intern
JM Financial Ltd - Mumbai, Maharashtra
July 2024 - November 2024
- Automated KYC processes using computer vision and deep learning, cutting processing time by 40%.
- Developed OCR solutions to enhance document verification efficiency by 30%.
- Analyzed large data to derive actionable insights, influencing strategic decisions and operational efficiency by 15%.
Backend Developer Intern
Kenmark ITAN Solutions - Mumbai, Maharashtra
December 2022 - April 2023
- Engineered robust APIs that boosted integration efficiency by 30% across multiple platforms.
- Implemented comprehensive QA protocols, enhancing overall system reliability by 20%.
- Revamped SQL and MySQL databases, significantly cutting query response times by 15%.
Featured Projects
AQI and Molecule Prediction
ViT-based model to predict AQI and molecular data from satellite and street view imagery. Optimized for accuracy and cloud scalability.
PyTorch MLFlow GCP ViT Computer Vision Multi-modal
Verbal-Vision
Lip reading model with 87% character-level accuracy. Deployed using TorchServe with optimizations for inference and training time.
PyTorch MLflow TorchServe Computer Vision NLP MLOps Docker
Image-Lingo
Image captioning model using attention mechanism with 91% accuracy. Integrated MLOps for model tracking and performance improvements.
PyTorch AWS MLflow Attention
Outreach-Ace
Gen-AI based resume analyzer and cold email generator with Langchain and LLMs. Includes a user-friendly Streamlit interface.
Langchain LLM ChromaDB Streamlit Gen-AI
Urban-Echo
Sound classification model trained on 8,000+ audio samples. Enhanced model accuracy and training using audio preprocessing and MLflow.
Audio PyTorch MLflow Docker
Latest Blog Posts
Agentic Patterns: The Building Blocks of Reliable AI Agents
ai machine-learning artificial-intelligence llm ai-agent
8/23/2025
Read More →
Automating Customer Support with AI: How Smart Bots Are Saving Brands (and Our Sanity)
artificial-intelligence ai automation customer-experience machine-learning
7/23/2025
Read More →
Machine Learning at Scale: Why PySpark MLlib Still Wins in 2025
data-engineering ai data-science machine-learning artificial-intelligence
7/11/2025
Read More →
How to Build Bulletproof Data Pipelines with PySpark That Actually Scale
artificial-intelligence data-engineering machine-learning python data-science
7/1/2025
Read More →
From Pandas to PySpark: My Journey into Big Data Processing
artificial-intelligence python big-data machine-learning data-science
6/9/2025
Read More →
Why Ethics in AI Matters: Tackling Bias and Building Fair Machine Learning Systems
ai ethics data-science machine-learning artificial-intelligence
6/4/2025
Read More →
Fine-Tuning LLMs in 2025: Techniques, Trade-offs, and Use Cases
machine-learning nlp llm data-science ai
5/20/2025
Read More →
Autonomous Horizons: How AI is Steering the Next Generation of Transportation
ai artificial-intelligence autonomous-cars computer-vision reinforcement-learning
5/18/2025
Read More →
The Silent Backbone: Why Traditional Machine Learning Still Matters in the AI Era
data-science ai machine-learning problem-solving innovation
4/21/2025
Read More →
Research & Publications
Examining ML Approaches for Early Diabetes Prediction
Authors: Yuval Mehta, Team
Conference: IEEE Asia Pacific Conference on Information Technology (APCIT) 2024
Abstract: This research explores multiple machine learning models for early diabetes prediction, highlighting key patterns in patient health data to aid in proactive healthcare measures.
View Publication →Contact Me
I'm always open to new opportunities and collaborations. Feel free to reach out!