Jun 2022 – Feb 2023 · 8 mos

• 𝟏. 𝐎𝐮𝐭𝐫𝐞𝐚𝐜𝐡 𝐓𝐞𝐚𝐦
• Description: A community to develop AI talent by providing training and education related to AI and Machine Learning via physical and virtual mediums
• ● Worked as an active member in a group of 25 people to create relevant content and project
• ● speaker for 3 different workshops on AI Education & Training in 3 three different educational institutions both physically and virtually

Nov 2021 – Feb 2023 · 1 yr 3 mos

• 𝟏. 𝐅𝐮𝐬𝐞𝐋𝐞𝐭𝐭𝐞𝐫
• Description: Community to improve the overall brand presence of Fusemachines in Nepal and also to create awareness among non-tech and the general public about AI
• ● Worked on content creation for the Monthly Newsletter about "The Art of Programming",
• "Tech Trivia", "Tech Quiz", "This Month in Tech", "Most-known Tools", and "Machine Learning Series" with the motive to AI Support Community
• ● Got high engagement from recipients.

Aug 2021 – Present · 4 yrs 9 mos

• Train Machine Learning models for Humans
• ● Led the ML team and developed domain-specific search-ranking systems using algorithms implemented from scratch, including BM25 and Word Mover’s Distance (WMD). Built a unified ranking pipeline with rare-token matching, custom WMD, and hybrid custom BM25 + SBERT for general search; genre-based relevance using custom WMD and hybrid custom BM25 + SBERT for movie search; cuisine-based relevance using custom WMD and hybrid custom BM25 + SBERT for restaurant search; and custom WMD with hybrid custom BM25 + SBERT for product search, achieving ~87% overall ranking accuracy. Optimized the swipe-based personalization workflow, reducing 8-swipe latency from 8–10 seconds to 4 seconds, and contributed to deploying the entire solution on AWS EKS for scalable production use.
• ● Developed and fine-tuned a conversational generative AI model using LlaMa, GPT-J, and GPT-Neo with the Hugging Face trainer API, leveraging medical domain knowledge to simulate a realistic doctor-patient interaction. Through comparative analysis, LlaMa emerged as the superior performer in this context.

Aug 2021 – Feb 2023 · 1 yr 6 mos

• 𝟏. 𝐑𝐞𝐬𝐮𝐦𝐞 𝐏𝐚𝐫𝐬𝐢𝐧𝐠 𝐚𝐧𝐝 𝐬𝐜𝐨𝐫𝐢𝐧𝐠
• Description: Platform that matches people with top AI companies in need of qualified engineers
• ● Lead the Resume Parser and Scoring project.
• ● Worked on end-to-end parsing and scoring pipeline that focuses on the development
• and deployment of NLP based services for job portals
• ● Identified whether the Document is Resume, JD, or Other Document.
• ● Segment the resume into different sections like Profile Segment, Objective Segment, Academic Segment, Experience Segment, and Skills Segment
• ● Fine tune BERT Transformer with spaCy 3
• ● Train a Joint Entities and Relation Extraction Classifier using BERT Transformer with spaCy 3
• ● Created an inference engine for information extraction and scoring.
• ● Shortlists the provided resume by generating scores based on the JD.
• ● Deployed the ML pipeline on AWS with Lambda, ECR, SQS, and S3 and put it Into the production
• 𝟐. 𝐙𝐞𝐧𝐝𝐞𝐬𝐤 𝐓𝐢𝐜𝐤𝐞𝐭 𝐀𝐮𝐭𝐨𝐦𝐚𝐭𝐢𝐨𝐧
• Description: The system automatically classifies the incoming tickets into one of the groups present in Zendesk based on the subject and subject of the inquiry(email/API/web). And the system suggests multiple macros based on their probability which are further classified into auto-suggestion(inquiries that will require manual intervention) and auto-response(inquiries that can be sent without manual intervention). The system also suggests multiple Frequently Asked Questions(FAQs) which are more similar to the inquiries. And the system also suggests a link to previous tickets where a similar inquiry was handled in case previous approaches don’t seem good.
• ● Worked on Email Automation system where for email classification SVM and Xgboost model were implemented and RTE models were implemented for macros suggestion.

Jan 2021 – Aug 2021 · 7 mos · Kathmandu, Bāgmatī, Nepal

• 1. Participate in the assessment process by invigilating exams.
• 2. Help develop, update and gather teaching material to support the development of the course curriculum.
• 3. Take on limited administrative responsibilities as requested by the Head of Department.
• 4. Provide in-person or email support to assist with student enquiries during the exam period.

Jul 2020 – Feb 2023 · 2 yrs 7 mos · Kathmandu, Bāgmatī, Nepal

• Grasp Coding is the official page handle by graspcoding.com. It offers flexible project-based on Python Code | AI Code | UI/UX Code also aimed to foster broad-based knowledge.