Samar Srivastava Email & Phone Number

Gurugram, Hr, In

Fully Remote

• Building POCs to bring ChatGPT enabled features to WiseStep CRM.• Handling data pipeline for integration of multiple VMS' with WiseStep CRM. • Stabilized in-house CV Parser which resulted in a 500% increase in CV processing speeds from 50CVs per minute to 300 CVs per minute. CV parser now runs at a speed of 1 million + CVs weekly. It predicts and extracts 30+ attributes from a CV like the industry of the candidate, education gaps, experience gaps, finding names from text via NER, etc.• Orchestrated strategy and logic for filtering out unique and duplicate active consultants across sources like emails, and job boards, marketed by different recruiters, using Levenshtein distance across multiple fields. This resulted in reducing duplicate candidates on the platform and achieving a higher degree of uniqueness, as a result, a recruiter can reach out to more unique candidates.• Optimized usage for Airflow for job orchestration, and scheduling to smoothen complex day-to-day data science activities like report generation, data extraction, etc.• Generate and manage data dashboards for the Customer Success team and stakeholders.Technology stack - python, airflow, ElasticSearch, Kibana, Apache Kafka, Docker, AWS SES, AWS Fargate, spacy, nltk, celery, flower.

Hyderabad, Telangana, India

• Responsible for developing job description parser (JD Parser) using nltk to scan through large amount of raw documents (emails) and extract information pertaining to job title, company name, compensations, locations, industry. JD Parser helped end users by attenuating job creation process by 70%. • Built multi platform, reusable, well-tested Boolean Query String Generator (BQSG) using Elastic Search, nltk, & spacy. BQSG is able to generate comprehensive boolean search queries just by taking in as input either of the following : Job Title, Job Description.• Developed industry classification model using pretrained BERT via transformers pipeline, & deployed as a fastAPI containerized via Docker. Response time on seqeunces with count of upto 2000 tokens was 81 milliseconds.• Performed Data scraping to collect data into Elasticsearch indices for educational institutes, professional degrees, and then performed pre-processing to build knowledge base for data products.• Developed Airflow DAGs to perform repetitive task like index cleaning, transportation pipelines to move data from ES to redis This ended up boosting teams productivity by a significant factor.• Responsible for end to end NLP pipeline starting from text pre-processing to packaging NLP tools for usage in production applications.• Tech Stack - python, fastAPI, elasticsearch, kibana, nltk, spacy, transformers, regex, airflow

Hyderabad, Telangana, India

- Currently working on information extraction/information retrieval using Natural Language Processing techniques- Applying NLP extensively to perform information extraction and fetching relevant attributes mentioned in client emails/attachments like job titles, skillsets, locations, compensation, organisations, etc.- Responsible for end to end NLP pipeline starting from text pre-processing to packaging NLP tools for usage in production applications.Tech Stack - Python, NLP, hugging face, tokenisers, spacy.

Gurgaon, Haryana, India

• Worked on data dashboard generation by analysing virtual assistants requests and response to detect anomaly in conversations and report malicious events.• Setup the pipeline for basic NLP preprocessing like text cleaning, tokenization, generating bag of words, evaluating n-grams. Evaluate similarity between text using cosine similarity, and much more.• Calculating tf-idf values over conversations, performing stemming, & lemmatization.• Working closely on production level deployment. • Used Kafka nodes as message brokers, capsulating modules into docker containers.• Used SQL Server 2017 as primary database and redis as in-memory for superfast data fetching.• Experimented with Uber AI's Plug & Play Language Model to induce personalities in text.• Responsible for development of a paraphrasing tool using Transformers.• Played major role in building online reputation of Scanta through helping hand in development of www.scanta.io website and web application based on machine learning and deployed on AWS. • Deployed 3 products on AWS using various services like EC2, API gateway, and AWS Lambda.Major tech stack - python3, huggingface tokenizers, redis, nltk, sklearn, Docker, T-SQL, flask, d3.js, and more.

Bachelor Of Technology - Btech, Computer Science

Selected projects: - Mooshy - A responsive buttons library written in pure CSS. - CryptoVisor - A windows application to analyse.

Full Stack Web Development Certification, Computer Software Engineering

Learned web technology and completed 4 courses from a coursework of more than 300 hours. Completed courses include : - Front End Libraries.

High School, High School Equivalence Certificate Program

Activities and Societies: Sports Team, assembly group