Bidyapati Pradhan

About Bidyapati Pradhan

With over 19 years of experience spanning IT Service Management, Machine Learning…

Experience

1. Machine Learning Engineer (Sr Staff)

   ServiceNow

   Dec 2020 — Present · Hyderabad, IN

   Current Work [2023-onwards]: Building LLM for ServiceNow:* Leading the team as Sr Staff Engineer and building various frameworks for LLM evaluation, sft/dpo data generation, training pipeline etc* Built in-house LLM evaluation framework(using LangChain) from scratch, to fill the gap from other frameworks. It is being used heavily by various internal teams to evaluate/benchmark checkpoints in all aspects/tasks. Currently supports 100+ industry benchmarks and 50+ internal benchmarks.* Identify model weaknesses and improve with model alignment.* Synthetically generated thinking tokens for Apriel-1.5-15b-Thinker model for Mid-training.* Built in-house framework(using LangGraph) for synthetic data generation, which is used for model fine tuning and model optimization.Past Work: 1. Worked with NLU team for ServiceNow NLU Engine development(Intent classification, Entity extraction).I was responsible for POC, development and support(NLU model issues, chatbot dialog flow analysis, latency issues in seconds to milisec response time) for NLU engine. Fixing NLU model in-house testing tools. Analysing customer NLU model with non-linear visualization techniques.[20••••22]2. Implemented OCR pipeline for document intelligence in Java for feasibility analysis, better performance and good user experience on latency.[2022]3. Built K-Mean and DBScan clustering pipeline in python.[2022]4. Built Similarity pipeline with embedding vector match and retrieval.[2023]Achievements:1. Analysed and fixed more than 30% customer cases(NLU/Chatbot/VA), most of them were performance oriented.[2021]2. Built K mean/DBSCAN clustering pipeline and Similarity pipeline in Python/Java from scratch. Submitted 3 patents in this area.[20••••23]3. Built LLM evaluation framework(low code, easy to extend/implement) for model evaluation from scratch.[20••••24]4. Built Synthetic Datagen framework(lowcode, reusable flows) for DPO/SFT data generation.[20••••25]

Education

• Veer Surendra Sai University Of Technology (VSSUT,Formerly UCE), Burla

  Bachelor of Engineering (B.E.), Electronics and Telecommunication Engg.

  2002 — 2006

• IIIT Hyderabad

  Foundational course, AI Machine Learning

  2018 — 2018

• Vedvyas College,Sambalpur University

  B.Sc, Physics Hons

  2001 — 2002

• Jawahar Lal Nehru College,JIEC,Ranchi

  I.Sc., Science

  1999 — 2001

Skills

• Java 8
• Service Activation
• Shell Scripting
• Oop
• Eclipse
• Java
• Unix
• Internet of Things
• Scripting(Shell,Batch,Perl)
• Unified Modeling Language (Uml)
• Akka
• C
• Clearcase
• Agile Methodologies
• Sdh/Sonet
• Web UI(Html,Css,Ext-Js,Sencha)
• Chatbot
• C++
• Containerization
• Ip Multicast
• Sql
• Linux
• Javascript
• Uml
• Object-Oriented Programming (Oop)
• Oracle
• Requirements Analysis
• C/C++ Stl
• Core Java
• Ipv4
• Perl
• Machine Learning
• Ason
• Deep Learning
• Debugging
• Design Patterns
• Optical Network