Building Intelligent Systems

Noise into products through
advanced ML, deepNET, and LLM modeling.
Mritunjay Pandey

Hello, I'm Mritunjay Pandey

AI/ML Engineer

I am an AI/ML Engineer with a B.Tech from NIT Jalandhar. Currently working as an Analyst at Aditya Birla Group. Previously, I worked at Namekart as an AI Engineer, at Invisible Technologies (Google DeepMind project) as an AI RLHF Agent (Contract), and at SEIL as a Machine Learning Engineer Intern, with a combined experience of around 1.2 years in applied AI and real-world problem solving.

Key Strengths

Machine learning, Deep Learning, Transformers & NLP, LLMs (GPT, BERT), Fine-tuning, RAG, PyTorch, Azure.

Skills

ML Core:
Supervised & Unsupervised Learning, Deep Learning, CNN, Transformers, NLP, Transfer Learning, Diffusion Models, Recommendation Systems, Hyperparameter Tuning
LLM & GenAI:
GPT, BERT, Fine-Tuning (LoRA, PEFT), RAG, Prompt Engineering, LangChain
Tools:
PyTorch, Hugging Face, Scikit-Learn, MLflow, Azure ML Studio, Docker, FastAPI
Data Stack:
ETL Pipelines, Feature Engineering, SQL, PySpark, Pandas, Matplotlib, Vector Databases (Pinecone)
Fundamentals:
Python, C++, Data Structures & Algorithms, Git / GitHub, Linear Algebra, Calculus, Probability

Education

Dr B R Ambedkar National Institute of Technology (NIT), Jalandhar
2021 - 2025
Bachelors of Technology - CGPA - 7.84
Certifications

Projects

Showcasing innovative AI/ML solutions and real-world applications

High Performance Two-Tower Recommendation System

High Performance Two-Tower Recommendation System

End-to-end semantic retrieval system using a fine-tuned MPNet Bi-Encoder with supervised contrastive learning for query–item alignment. Implements FAISS HNSW approximate nearest neighbor indexing for O(log N) vector search and ONNX Runtime Int8 quantization for 5× faster CPU inference. Deployed as a real-time microservice with integrated A/B latency benchmarking.

LangChainLlamaFastAPI
View Project
Agent Portfolio and Valuation Advisor (APVA)

Agent Portfolio and Valuation Advisor (APVA)

LangChain-powered agentic framework running a multi-LLM pipeline (Llama-3.3-70B, Nemotron-49B) with streaming inference, retry logic, and 5+ API integrations using rate limits and graceful fallbacks for 99.9% uptime. Developed an LLM-driven multi-tier pricing engine generating six strategic price points with structured JSON parsing, rationales, buyer personas, and hold-time optimization.

LangChainLlamaFastAPI
View Project
Multi-Agent LinkedIn Optimization System

Multi-Agent LinkedIn Optimization System

Developed single-agent AI system for profile analysis (0-100 scoring), job fit assessment, content optimization (ATS-friendly), and career guidance using structured prompt engineering. Implemented a conversational AI coach with session memory, JSON-based data export, and a LinkedIn-style real-time feedback UI in Streamlit.

LangChainConversational AI
View Project
Medicine Recommendation System

Medicine Recommendation System

Emergency-ready ML assistant that personalizes drug recommendations from symptoms, vitals, and EHRs while keeping data private. Backed by Kaggle clinical datasets for trustworthy suggestions in critical care.

Jupyterscikit-learnPandas
View Project
Virtual Painting Studio

Virtual Painting Studio

Gesture-driven drawing surface powered by OpenCV + MediaPipe that follows index-finger paths, toggles colors/eraser with multi-finger poses, and adjusts brush thickness through thumb-index distance.

OpenCVMediaPipeFlask
View Project

Experience

Aug 2025 ----- - Present

Business Analyst

Aditya Birla Group - Grasim

Mumbai, India

  • ▹Converted SAP sales data into Excel and Power BI dashboards for MIS insights.
  • ▹Daily production analysis through Birla Assure to detect deviations across 3200 TPD plants.
  • ▹Supported Business Development by analyzing home textiles market and assisting 200 kg Modal hub development.
SAPExcelPower BIData Analysis
Jan 2025 ------ Aug 2025

AI Engineer

Namekart Pvt. Ltd.

Noida, India

  • ▹Trained Transformer + MLP on 767K+ domains for price prediction (L1 $270, L2 $500).
  • ▹Built DistilBERT + MLP with SE-block fusion and Huber Loss.
  • ▹Automated lead discovery and classification using LangChain and LinkedIn scraping.
  • ▹Deployed on Azure ML with confidence-based price range outputs.
PyTorchHugging FaceAzureTransformerMLP
Sep 2024 ------ Jan 2025

AI Supervised Fine tuning Agent

Invisible Technologies

Remote, USA

  • ▹Created 700+ gold-standard triads for Python and STEM SFT and RLHF training.
  • ▹Achieved 98% QC accuracy; promoted to Reviewer for LLM dataset validation.
  • ▹Evaluated 2,000+ prompts using SOPs and Human-in-the-Loop review to reduce hallucinations.
PythonChain-of-ThoughtRLHFPrompt Engineering
May 2024 ------ Aug 2024

Machine Learning Engineer

Statcon Electronics India Limited (SEIL)

Delhi, India

  • ▹Built CNN-LSTM SoH model with 96% accuracy and RMSE 0.02.
  • ▹Processed 10K+ cycles and predicted failures 50 steps ahead.
  • ▹Deployed Streamlit demo with sub-150 ms inference.
Seq2Seq ModellingStreamlitTime Series Analysis
Let's Talk