Prashant J.

About Me

Prashant is an AI/ML Engineer with 6+ years of experience developing AI-powered solutions using Large Language Models, machine learning, and conversational AI. Proven track record of building scalable systems that increase business metrics by up to 18%. Expert in open source LLMs fine-tuning, RAG implementation, and real-time AI deployment. Successfully led development of interview platforms, resume parsers, and recommendation engines serving millions of users.

AI, ML & LLM

VectorDB Apache Airflow Machine Learning Deep Learning GPT-4.0 Pytorch Conversational AI embeddings NLP LangChain VLLM Agentic AI LLaMA 3.1 LLM

Backend

Python REST APIs

DevOps

AWS Sagemaker Api Gateway CodeDeploy Docker AWS Lambda AWS EC2

Other

QLora Word2Vec Cross-functional Team Management AB-Testing Langraph Pandas time series forecasting Clustering Algorithms Deepgram Tensorflow Trans-E RAG Siamese Network Vector db Named Entity Recognition Numpy Prompt Engineering Team Leadership WebSockets Keras MCP LoRa Quantization Fine-tuning knowledge graph Stakeholder management Data Science Hugging Face

Work history

Deloitte
Deloitte
Lead Data Scientist
2023 - Present (2 years)
Gurgaon, India
  • Leading a team of engineers and data scientists to develop an AI-powered interview platform using LLaMA 3.1, fine-tuned with LoRA and QLoRA, reducing client interview scheduling time by 60% and improving candidate experience scores by 85%.

  • Leading the development of an intelligent resume parser using fine-tuned LLaMA2 13B model with parameter-efficient fine-tuning, achieving 91% accuracy in structured data extraction.

  • Developed Job Description to resume matching engine using embeddings and RAG architecture, increasing successful placement rates by 35% and reducing time-to-fill by 45%.

  • Implemented semantic search with Qdrant vector database and multi-query decomposition for enhanced relevance matching.

  • Spearheaded the Auto-Calling Engine project, increasing candidate screening efficiency by 300% and lowering cost-per-hire by 40%.

LLaMA 3.1 LoRa QLora VLLM Deepgram RAG Qdrant embeddings Data ScienceAI/ML Data Extraction Parsers Vector Databases Generative AI Natural Language Processing (NLP) Large Language Models (LLMs) Data pipelinesLangChain TransformersAgentic AI PythonRetrieval-augmented Generation (RAG) MCP
PharmEasy
PharmEasy
Senior Data Scientist
2021 - 2023 (2 years)
Bangalore, India
  • Developed a comprehensive recommendation engine platform combining collaborative filtering and NLP-based matching systems, resulting in an 18% increase in Click Through Rate and 9% increase in Add To Cart metrics across the platform through rigorous A/B testing.

  • Developed an intelligent medicine name normalization system using Siamese Network for automated product matching with PharmEasy golden master database, increasing automapping percentage by 12% and reducing the Data Governance team's manual workload by 200+ hours weekly.

  • Implemented an NLP solution that handles product name variations, spelling errors, and brand synonyms, enabling seamless product catalog management and improving inventory accuracy by 25%.

  • Trained Word2Vec on the transaction data to generate most frequently used items with respect to a given item.

Siamese Network Data ScienceNatural Language Processing (NLP) A/B TestingAWSSocket Programming Collaborative Filtering SQLTransformersMongoDBRecommendation Systems Word2Vec
Accenture
Accenture
Data Scientist
2019 - 2022 (3 years)
Gurgaon, India
  • Built a comprehensive talent intelligence platform using Named Entity Recognition and knowledge graph algorithms to extract skills from resumes, create skill ontologies with Word2Vec embeddings, and develop reskilling frameworks using Trans-E algorithms, reducing bench size by 30% and improving internal mobility by 45%.

  • Developed a predictive analytics suite for workforce optimization including time series forecasting models for demand-supply-attrition gaps and employee exit analysis using clustering algorithms to identify retention drivers, enabling proactive talent planning and reducing attrition costs by 20% across service lines.

  • Developed a Deep Learning-based model (NER) to extract skills and job title from resumes and job descriptions.

  • Developed a keyword summarization module using Selenium (extracted 100 pages of a particular keyword), TFIDF (extracting important words), language model (generating embeddings from keywords-pretrained model), clustering using generated embeddings.

  • Worked on dashboard development using d3.js and JavaScript.

Named Entity Recognition Trans-E TransformersKnowledge Graphs Predictive AnalyticsData ScienceTime Series AnalysisWord2Vec Clustering Algorithms Machine LearningKerasApache SparkAWSAWS Lambda PythonDashboard Development d3.jsJavaScriptDeep LearningSeleniumWorkforce Management (WFM)
EDGE
EDGE
Data Scientist
2018 - 2019 (1 year)
Bangalore, India
  • Improved user experience through Data Science to deliver compelling value in profile manager.

  • Worked on title normalization to improve the quality of the search & match product.

  • Worked on an NLP-based model to segment different sections from resumes (personal information, projects, skills, certifications).

  • Tagged ~60K resumes and trained classifiers on top of it.

NLPData ScienceDocument Parsing Parsers UX

Education

Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization | Neural Networks and Deep Learning | Structuring Machine Learning Projects
Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization | Neural Networks and Deep Learning | Structuring Machine Learning Projects
Coursera
2017 - 2017
B.Tech ECE
B.Tech ECE
The LNM Institute of Information Technology - India
2014 - 2018 (4 years)