Jun 2025 – Present · 11 mos · Remote

• Architected Triple M BrainBox, a SOTA EdTech platform generating adaptive NEET/JEE quizzes via
• RAG with vector search and GPT-4.1-mini integration.
• Built a Pub/Sub event-driven system on GCP with containerized FastAPI services, Next.js + Tailwind
• frontends. Implemented AI-driven reviews, caching and analytics, powering 1,000+ DAUs

Jan 2025 – Jun 2025 · 5 mos · Remote

• Contributing to the startup’s growth by delivering ML, Generative AI, Data Analytics, and iOS solutions across various projects.
• Key Contributions:
• Developing an iOS application integrating complex ML models for real-time image recognition, NLP-based text processing, and predictive analytics.
• Implemented an on-device ML pipeline using Core ML, TensorFlow Lite, and Metal Performance Shaders (MPS) to ensure optimized performance.
• Built a Generative AI-powered chatbot using RAG (Retrieval-Augmented Generation) architecture with LangChain, Pinecone, and OpenAI APIs for intelligent user interactions.
• Designed data analytics dashboards leveraging Python (Pandas, NumPy, Matplotlib), Apache Spark, and Power BI, providing actionable insights to business teams.
• Integrated RESTful APIs and GraphQL endpoints into the iOS app using Swift, Combine, and Alamofire, ensuring seamless backend communication.
• Optimized cloud infrastructure for AI workloads using AWS Lambda, S3, and DynamoDB, improving efficiency and scalability.

Sep 2024 – Present · 1 yr 8 mos · Remote

• As a freelance developer, I’ve delivered high-quality tech solutions across multiple domains. Notably, I developed the official website for Triple M Group, the largest educational group in Hoshiarpur, Punjab.
• Beyond web development, I specialize in ML and Generative AI solutions. I’ve worked with a fintech client to implement ML-driven solutions and have built RAG-based chatbots for apps and websites, enhancing customer interaction and automation.

Jun 2024 – Jul 2024 · 1 mo

• Building and deploying Generative AI applications leveraging LLMs (Large Language Models) for chatbots, text summarization, and content generation using OpenAI GPT, LangChain, and Pinecone.
• Developed a RAG-based chatbot for web and mobile apps using Vector Databases (Weaviate, Pinecone), LangChain, and FastAPI, enabling intelligent, context-aware responses.
• Designed end-to-end ML pipelines for data preprocessing, model training, and deployment using TensorFlow, PyTorch, Hugging Face, and MLflow.
• Deployed ML models using Docker, Kubernetes, AWS Lambda, and FastAPI, ensuring efficient scaling and real-time inference.
• Implemented MLOps workflows using CI/CD pipelines (GitHub Actions, Docker, Kubernetes) for seamless model versioning, monitoring, and retraining.
• Optimized inference performance by quantizing models with ONNX Runtime, TensorRT, and Hugging Face Optimum, improving efficiency for production deployments.
• Developed data analytics pipelines for AI-driven insights using Apache Spark, Pandas, and Snowflake, helping improve business decision-making.

Jan 2024 – Dec 2024 · 11 mos · Remote

• Training and fine-tuning LLMs (Large Language Models) using RLHF methodologies, incorporating human feedback loops to optimize model behavior.
• Curating, labeling, and evaluating training datasets, ensuring AI-generated responses align with human intent, ethical considerations, and factual accuracy.
• Implementing reward modeling using Proximal Policy Optimization (PPO) and reinforcement learning techniques with Hugging Face’s TRL, RLlib, and DeepSpeed.
• Designed custom preference ranking systems, leveraging pairwise comparison and Elo-based ranking to guide model fine-tuning.
• Developed scalable training pipelines using PyTorch, TensorFlow, Hugging Face Transformers, and DeepSpeed, optimizing efficiency and computational cost.
• Conducted bias detection and mitigation using AI fairness techniques (SHAP, Fairlearn, LIME) to ensure ethical AI behavior.
• Worked on LLM evaluation frameworks, using BLEU, ROUGE, perplexity scores, and human-rated assessments to benchmark model improvements.

Jun 2023 – Jul 2023 · 1 mo

• Developed and fine-tuned ML models for predictive analytics, anomaly detection, and recommendation systems using TensorFlow, PyTorch, and Scikit-learn.
• Built an end-to-end ML pipeline for data collection, feature engineering, model training, and deployment, leveraging Apache Airflow, MLflow, and Kubernetes for automation.
• Designed a real-time fraud detection system for a fintech application, using XGBoost, LightGBM, and anomaly detection algorithms, deployed via FastAPI and Docker.
• Created automated retraining pipelines with CI/CD (GitHub Actions, DVC, and MLflow) to ensure models stay updated with the latest data.
• Engineered data pipelines using Apache Spark and Snowflake, processing large-scale datasets for model training and business intelligence.
• Conducted A/B testing and model evaluation using SHAP (Explainable AI), ROC-AUC, and Precision-Recall metrics to ensure model fairness and robustness.