{'contact_info': {'email': '[email protected]',
'phone': '9503554884',
'websites': ['https://www.linkedin.com/in/swarupd/']},
'education': {'college': ['Liverpool John Moores University'],
'degree': ['MSc Data Science',
'Bachelor of Engineering Computer (Year 2018']},
'metadata': {'job_pk': 12,
'language-code': 'en',
'language-confidence': 0.9999960436887447,
'remark': 'success',
'status': 'succ'},
'personal': {'location': 'Pune, India', 'name': 'Swarup Das'},
'skills': ['Matplotlib',
'Postgresql',
'Xgboost',
'Metrics',
'Segmentation',
'Pandas',
'Api',
'Startup',
'System',
'Governance',
'Analysis',
'Keras',
'Python',
'Java',
'Engineering',
'Engagement',
'Mysql',
'Migration',
'Forecasting',
'Updates',
'Flask',
'Sql',
'Analytics',
'Aws',
'Cloud',
'Seaborn',
'Sanic',
'Warehouse',
'Tensorflow',
'Docker',
'Etl',
'Hive',
'Algorithms',
'Numpy',
'Pyspark',
'Auditing',
'Queries',
'Testing',
'Technical',
'Spark',
'Hypothesis'],
'work_exp': [{'designation': 'Lead Machine Learning Engineer (Fintech)'},
{'designation': 'Data Scientist'},
{'designation': 'Software Engineer Trainee'}]}
The deliverable should include the following:
- Definition of the accuracy of the resume parser → How do we evaluate success? ->Parsing resumes is typically a classification problem , to be more precise it is a multiclass classification problem. we can use metrics like F1 Score, Precision,Recall, ROC- AUC to evaluate the performance . While training our model, the best performing model is the one which had the highest F1 Score.
E # LOSS TRANS... LOSS NER ENTS_F ENTS_P ENTS_R SCORE
--- ------ ------------- -------- ------ ------ ------ ------
0 0 9238.97 1544.76 0.20 0.10 3.45 0.00
3 200 199317.57 66695.30 32.15 39.21 27.24 0.32
6 400 58940.22 23063.40 55.83 59.72 52.41 0.56
10 600 11210.37 20986.48 54.71 52.11 57.59 0.55
13 800 8272.94 16835.60 44.18 70.99 32.07 0.44
16 1000 11448.79 16229.70 53.76 48.03 61.03 0.54
20 1200 6256.43 16764.19 55.08 64.43 48.10 0.55
23 1400 6140.25 14505.28 56.56 61.41 52.41 0.57
26 1600 6545.44 14061.78 60.71 70.55 53.28 0.61 ---- Best Model
30 1800 586.02 14062.91 58.19 58.80 57.59 0.58
33 2000 4867.82 12252.47 54.72 69.79 45.00 0.55
36 2200 44495.01 12424.94 60.04 65.84 55.17 0.60
40 2400 831.39 12326.80 60.22 63.95 56.90 0.60
43 2600 8369.49 10585.22 58.05 68.54 50.34 0.58
46 2800 731.13 10233.91 56.37 67.39 48.45 0.56
50 3000 254.50 10085.33 58.71 61.24 56.38 0.59
53 3200 15952.13 8614.01 56.12 66.35 48.62 0.56
-
The architecture of the Resume Parser (high-level) - How does it work? (e.g. holistic approach, deep learning models, etc.)
- The Parser first tries to convert pdf to text
- then it tries to detect the langugae of the doc.
- After that it performs extraction/prediction using Regex, NER and custom spacy model
- Post that the results are saved into the SQLite DB
-
Milestones to complete the project with an estimated timeline
This is an initial implementation , would refine more depending on the feedback
-
Basic implementation of the Resume Parser
- Simple API service (e.g. FastAPI service)
- Resume Parser (e.g. Python)
{'Education': ['MSc Data Science (Year 2021), Liverpool John Moores '
'University, Grade : Distinction',
'Bachelor of Engineering Computer (Year 2018), Pune University, '
'Grade : Distinction'],
'Skills': ['Machine Learning',
'MLOps',
'Data Engineering',
'ML Algorithms',
'Libraries',
'Big Data/DataBases',
'Cloud',
'Languages'],
'Work experience': [{'company name': 'OneCard by FPL Technologies, Pune, '
'India',
'designation': 'Lead Machine Learning Engineer',
'end-date': 'Present',
'start-date': 'March 2020',
'summary': 'Built amazing products like OneScore (20 '
'Million + Users) & OneCard (~ 2 Million '
'Users). Led the team of engineers to create '
'Data Lake, ML Workbenches & Analytics '
'Platform.'},
{'company name': 'Rule 14 LLC, Pune, India',
'designation': 'Data Scientist',
'end-date': 'Feb 2020',
'start-date': 'Jan 2019',
'summary': 'First hire of the data science team and was '
'also responsible for team building and '
'heading initial client projects'},
{'company name': 'Searce LLC, Pune, India',
'designation': 'Software Engineer Trainee',
'end-date': 'Dec 2018',
'start-date': 'July 2018',
'summary': 'Worked on Image captioning (using VGG16), '
'Model Blending and learnt best practices '
'like CI/CD, Canary Deployment, Test Driven '
'Development etc'}]}
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent