Vedant K. - AI Engineer

Pune, India

Experience

Jul 2025 - Present

6 months

Pune, India

Led end-to-end client AI projects from discovery to deployment, delivering NLP, computer vision, and automation solutions that drove measurable business impact.
Designed and deployed LLM-powered agentic applications and chatbots, applying prompt engineering, evaluation, and LLMOps for scalable performance.
Developed production-ready code and deployed solutions on cloud platforms, ensuring reliability, scalability, and cost efficiency.

Jan 2025 - Dec 2025

1 year

Developed multi-agent RAG system using LangChain and LangGraph to orchestrate intelligent workflows, integrating Hugging Face transformer models with SQL databases for hybrid retrieval from structured and unstructured data sources.
Engineered a natural language-to-SQL translation pipeline leveraging fine-tuned language models to convert user queries into optimized database operations, achieving sub-second response times through strategic caching and GPU-accelerated inference.
Built intelligent agent architecture with specialized roles (Query Router, Document Retriever, SQL Analyst) that autonomously determines optimal data sources and performs iterative refinement, improving response accuracy by implementing conversation memory and explainable AI features.

Jan 2024 - Jun 2025

1 year 6 months

Chicago, United States

Developed retrieval-augmented generation (RAG) systems using LangChain and LangGraph frameworks, applying statistical analysis and evaluation methods to reach 92% accuracy on unstructured data tasks.
Built scalable data pipelines to transform clinical documentation into analysis-ready datasets, using validation and pre-processing techniques that cut preparation time by 60% and boosted downstream model performance.
Designed evaluation frameworks that applied performance metrics, error analysis, and fine-tuning approaches to iteratively improve model accuracy while reducing retraining overhead.
Collaborated with cross-functional teams to integrate LLM-powered AI solutions into real-world processes.

Jan 2024 - Dec 2024

1 year

Developed a specialized version of LLaMA 2 (7B) that converts natural language descriptions into executable code with high precision.
Implemented efficient data pipelines to process 500K text-code pairs using Parameter Efficient Fine Tuning (PEFT) and 4-bit quantization techniques.
Engineered the training infrastructure using Nvidia CUDA for parallel processing and optimized memory usage.
Achieved significant reduction in computational requirements and 3x faster training time compared to CPU-based approaches.

Jan 2022 - Dec 2022

1 year

Developed an AI system that accurately identifies human emotions from speech patterns for enhanced user interaction analysis.
Designed custom data pipelines for acoustic feature extraction using TensorFlow and Keras, incorporating LSTM and CNN architectures for temporal and spatial pattern recognition.
Implemented a hybrid model combining deep learning with linear regressors to capture complex emotional cues in audio signals.
Achieved 82% accuracy in emotion classification, representing a 15% improvement over existing models.

May 2021 - Oct 2022

1 year 6 months

Mumbai, India

Built NLP projects in Python and R to analyze and extract key information from text, helping reduce manual review work by 40%.
Assisted in integrating machine learning models into data workflows, improving processing efficiency 30% and making predictions more reliable.
Created data analysis tools to turn complex datasets into clear insights, supporting team projects and saving costs on outside analytics work (approx. $150K annually).