• Domain based Article Generation for Sports & betting companies based on brand style & language. Sports covered NFL, NBA. Use Generative AI for custom fine-tuning on styles and sport scenarios. Lead the team and product from scratch and delivered.
• Repair Manual RAG chatbot, automobile understanding knowledge bots, Document extraction API's for auto-insurance sector.
• Built LLM chatbots recruiter chatbot, hardware faucet stores software with RAG
• Built article generation software with LLMs for sports media companies. Understood the domain and scenarios of the product. Proactively worked as a Gen AI product person and developed the LLM based product.
• Built recursive candidate validation software for recruiter analytics involving GenAI and scraping for client. This will effectively extract relevant data of candidates from web and do a personality analysis.
• Built automatic SRS document maker, refiner, troubleshooter (involves question extractor, tech stack, requirements, team mix, complexity) and Launch first estimator with LLMs
• Fine-tuned custom LLMs on custom data for text to sql queries, tech specific code retriever (used pangucoder method), recipe maker, cricket analyst with PEFT. Worked on GPU nodes for doing pre-training and instruction fine-tuning on LLMs. Have good understanding of quantization, GGML/UF models, ollama frameworks, llama-index, langchain, advanced RAG. Have also create custom datasets with the idea of WizardLLM.
• Have implemented customer support dialogue chatbot on custom data. Used LLMs, OpenAI GPT3.5, langchain, python, fastapi, aws, prompt engineering, pinecone for the client Have built ad recommendation pipelines and models on big data. Have used with aws sagemaker, aws EMR,
• Aws athena, spark, python, bigdata processing, machine learning, CI/CD, jenkins, MLFlow, SQL, Athena SQL, pyspark, pytorch, aws sagemaker, extensive datanalaysis, sagemaker pipelines, deployment, dockerization - model buiding & data pipelines for the client

• I worked on building prediction models for First Party Audience. Some of the prediction models are Homeowners prediction, business travelers, PSA users prediction, T2D users prediction, Asthma users prediction, etc., for IOS, android and Web based users. Used features based on MSA, Social determinants, behavioural data, etc. Leveraged using Machine learning, Deep learning for training models. Have done a lot of EDA to find right features and soft labels. Have worked on building effective evaluation metrics along with using advanced probability based correlation on MSA data.
• Worked on building applications using LLMs, vector databases like weaviate/elastic, RAG. Have built & evaluated search functionalities, content generation & relevance, RAG chatbots on filtered contents. (multiple usecases). Applied Generative AI on Weather article relevancy checks with multiple data points.
• Along with using existing LLMs like openai GPT4, mistral, etc., I have finetuned LLMs like gemma for content evaluation. I have also evaluated MTEB embedding models for articles search usecases.
• Worked on Large scale data modelling of around 400 features and 600M records. On daily basis, the client get 3M new records into the database. Worked with Huggingface, RAG, weaviate, fastapi, ragas, PySpark, EMR, Sagemaker pipelines, python, PyTorch, sklearn, Aws Athena, gin configs. fastapi. dockerization & deployment with CI/CD & jenkins. Built Gen AI frameworks & large data pipelines

• Implemented Table Detection and Table Structure. Detection on banking documents using Cascade TabNet, Table transformer and Layout LM. (LLM)
• LLM's Implemented Native PDF Extraction using fitz and paddle OCR.
• This helped the company to reduce costs on 3rd party APIs for information extraction. I have generated models with 94% accuracy on bank documents equivalent to 3rd party APIs.

Worked on building AI features using GPT- 3 for ATS. Resume parser, requirement parser and search engine.

• Handled 4 batches till now (DLCP-PGP August 2021 batch, PGP-ML December 2021 batch)
• + Course given to Jain University PGDM in collaboration with Great Learning.
• + Course giving to IIIT-H PGDM in collaboration with Great Learning.
• (Python, SQL, DSA, Data Visualization, ML, DL, NLP, Recommendation Systems)

• Have mentored groups of students from 8 batches (DCS27, DSC23, MLC21, DSC14, DSC11, MLC8, DSC11 career phase, MLC8 career phase)
• Mentoring & Training Students into Data Science, NLP
• Building Professional Confidence to students / mentees
• Giving Live sessions, group sessions, concept wise recordings, seminars, webinars on various AI topics
• Doubt resolution, grading
• Career guidance

• Contributed to build AI platform from scratch. My use-cases & work help company reduce costs and increase Customer acquisition.
• Worked with GPT-3, Open AI, BigBert, LLM (large language models) as well Core Competencies NLP, Python, Devops ,Data Modeling Statistics, Data Visualization, Business Analytics, Deployment.
• Leadership: Team Building, Team Training
• Product Development & Project Management (ATS Staffing & Recruitment Analytics, HR Analytics, CRM Analytics)
• Analysis: Analyzing Requirements by coordinating with Clients and Product managers based on possibility, need and Usage.
• Deployment: AI models & server Deployment, Maintenance. Dockerization of our product services.
• Analytics: Market, Customer and Business Analytics - Choosing next use cases.
• Recommendation Systems, Text Similarity, Document Similarity, News detection - Data Architecture And Design, Designing Approach, Building custom algorithms and Custom frameworks.
• Pre-screening Chatbots, Employer Chatbots. Research: Computer Vision based Video Interviews via Deepface - emotion capture, video analysis, audio analysis, Face verification, speech to text, etc.

Achievements:

• Built Resume Parser & JD parser by training India, US and Singapore based resumes/jds considering both technical and non-technical. (A unique method for data extraction with great tools are also part of this and OCR)
• Achieved 84% F1-score on combined parsed fields. Reduced the 3rd party api costs for the company. Currently Parsers can handle parsing of 2L resumes daily via api and data migration together. Parsers could rapidly work with a response time of 200 ms on average.
• Implemented job scraping engine and analytics from over 65 job boards and prime vendors scraping 1000 jobs every 5 mins.
• Built pre-screening chatbot that saves users time by reducing follow-ups and time management. Effectively used feature by over 1600 users on daily basis for capturing data of around 10000 candidates.
• Pre-screening Chatbots, Employer Chatbots
• Research: Computer Vision based Video Interviews via Deepface - emotion capture, video analysis, audio analysis, Face verification, speech to text, etc. Pipelines: Client Data Migration - Automation.

Projects:

• Resume Parser, Job Description Parser, Email Parser, Document Parser, LCA Parser - NLP Information Extraction, Intent detection, Topic modelling, Custom Entity extraction - Email Classification, Document Classification, Domain Classification, Resume Classification - NLP, Document conversions, Automatic Search Query builder.
• Implemented Keyword Search and Semantic Search for our Candidate Ranking System - (accurate and fast indexing search) - (Advanced search capabilities and filters) Implemented Candidate Matching & Ranking, Candidate

Analytics

• Built Scraping Engines for our use case to grab jobs, candidate profiles, recruitment data/news - Leveraged GPU For training on high-end algorithms fastly and accurately Email verification and IMAP
• Computer Vision for OCR related information extraction from images, pdf parser, xml parser, box detection, entity detection and recognition.
• Data Mining, Data tagging, Data Analysis, Feature Selection & Model Selection, Model Building, Model Validation, Model threshold validation, log analysis.

Skills used: Dockerization - docker images, docker containers, docker compose, aws app services, s3, azure app services, server performance monitoring, cron jobs, Python, flask, gunicorn, Linux/Ubuntu, APACHE SOLR CLOCLOUD, MS SQL, solr faceting, solr mlt, Banana dashboard, NLP - Information Extraction, NER, Apache Solr, Solr MLT. (solr more like this), Solr Indexing, Spark, topic modelling, SQL, Python, multi processing, Machine Learning, scikit- learn/sklearn, pandas, numpy, NLP, NER, Information Extraction, CRF, Word2vec, Fastext, Doc2vec, WMD words movers distance, spacy, gensim, pattern, textblob, Neural Networks, Deep Learning, ANN, CNN, RNN, LSTM, BILSTM, BERT, box detection, Web Scraping, scrapy, scrapy cloud, crawlera, selenium, splash, spiders, docker, linux/Ubuntu.

• Worked on Audio codec analysis (mainly for G711, Speex, Opus, AMR, Flac) for our ASR engine with respect to frame Sizes, bit rates, sample rates, audio quality, etc., for our Products / environments, requirements and end-to-end Implementations.
• Worked on NLP chatbot using Google Dialogflow & RASA - Streaming audio and integrated it with our product using GRPC CPP / Python Application. Implemented it convert our Command based product to free flowing human language Product !
• Log data analysis, time based analysis for Nuance Speech engines. Training Speech system with phenomes inputs.
• Text mining / analysis using Word2vec model, CRF model,LSTM - CNN models.

Recent implementation: Text mining / analysis using Word2vec model, Words Movers distance algorithm, Fastext model, CRF model, LSTM - CNN models, Stanford Dependency parsers. Also implemented BIDAF (Bidirectional attention flow model) of Stanford Squad dataset

Responsibilities-

• Developed Chatbots using NLP / NLU - GATE text Engineering ? Extracting intents/entities from text (randomized sentences written in different ways), scoring, Making synonym module, antonym module, randomized Time and date extraction module, phrase extraction Modules, etc. Integrated all to my chatbot framework. Worked on this for Banking, Telecom, IT-Support, Finance Domains.
• Implemented MIS Chatbot, which in summary stores Context of previous queries of the user and also shows Visualization for the finance related queries from the Database. (Here, we find out the intent of randomized Finance query, make an SQL query using that intent and Extract data from database and show visualization to user. Based on the user query, our chatbot can decide to reply or Show data or show visualizations)
• Made and used vast number of modules using GATE at Senseforth for intent / phrases identification from query Texts and documents.
• Also worked with word2vec, fastext, WMD, dependency Parser for text analysis using Python.

• Worked on Candidate Rankings and Candidate Comparison use case using Machine learning Random forest for our client data. Created a Tableau visualization for it.
• Worked on clustering use-case for telematic data to Analyze achievable profits for the client. Worked on telecom Data to analyze achievable profits for the client. Prescriptive analytics. - Worked on Email Sentimental Analysis problem for my client using Bag-of-words and Random forest model (Text Classification)
• Worked on fraudulent claim predictions, client subscription. Analysis. - predictive analytics. - Worked on Chiller plants Maintenance data using Linear Regression Analysis, Exploratory Analysis, Association Analysis to get the insights From data using R, Python and thereby detecting spikes of The plant.- Descriptive analytics, EDA.

Personal Implementations:

• First Chat-bot : Made a simple FAQ Chatbot using cosine similarity, python- NLTK, pandas, numpy libraries. Also used web scraping / crawling to get live data.
• Solved Loan prediction problem using Decision Tree/Random forest algorithms.
• Extracted & Worked of twitter data, facebook data to do exploratory analysis.

Software Engineer- (June 2015- 2016)

• Worked on solving Cross-Site Scripting Vulnerabilities using Cold Fusion technology.
• Worked on developing and enhancing the features of web applications using ASP.NET 2016
• Worked on customer data and its intricacies using MS CRM and MS NAV tools. -2016
• Learnt python (ds,oops), dbms, integration of py & dbms
• .Net - C#(basic & advanced), webforms, ADO.NET, ASP.NET MVC etc.
• Worked on developing python applications for airline industry.
• Worked on developing ASP.NET (MVC, webforms also) based applications for insurance Industry.