Bhoomika R Hegde

1. 2026

   Transaction Graph-Based Predictive Hurdle Model for Credit Scoring in DeFi Lending Protocols

   B. R. Hegde et al.  ·  IJDSA

   PUBLISHED
2. 2026

   Discrete Representation Learning: A Review

   B. R. Hegde et al.  ·  ICMLC

   TO BE PUBLISHED

• Systems · ML · 2024

  Dog Activity Recognition System

  Built a system to study dog activity from video and biosignal data. It uses Python, Kafka, PostgreSQL, FastAPI, and Docker to keep video frames and sensor streams aligned in time. A CLIP-based inference service supports zero-shot activity recognition, with logs, retries, and dead-letter queues for safer failure handling.

  • Python
  • Kafka
  • FastAPI
  • Docker
  • CLIP

  View on GitHub

• ML · Research · 2024

  Image Representation via ViT + Product Quantisation

  Worked on a new image representation method after finding limits in VQ-VAE2. The project combines Vision Transformers with product quantisation to learn compact image codes. Results were checked using reconstruction quality, SNR, and SSIM, and the companion paper is accepted for ICMLC 2026.

  • PyTorch
  • ViT
  • VQ-VAE
  • SSIM

  View on Google colab

• Backend · Systems · 2024

  Relational Knowledge Graph for Question Answering on Relational Databases

  Built a question-answering layer for relational databases. The system reads the database schema, builds a relational knowledge graph, and uses that structure to plan queries and trace how an answer was found.

  • Python
  • SQL
  • Graph Theory

  View on GitHub

• Backend · Architecture · 2024

  Academic Management System

  Built a university portal using separate backend services for scheduling, feedback, and user management. Each service had its own API contract, making it easier to update one part without touching the rest. The system was containerised with Docker.

  • Docker
  • REST
  • DDD
  • Microservices

  View on GitHub

No projects match this filter yet.

1. Jul 2026 — Present WLDD

   AIML Engineer Intern

   • Building a backend microservice that collects creator profiles from Instagram and Twitter, stores clean metadata, and adds AI-generated summaries and embeddings.
   • Designed a two-stage async pipeline for discovery and enrichment, using BullMQ workers and MongoDB so creator records can be reprocessed without blocking new ingestion.
   • Implementing platform-specific scrapers and content-analysis flows that support internal apps like DMS, SOLO, meme'd, and a future AI creator-search chatbot.
2. Feb — Jun 2026 Yugasys

   AI Engineer Intern

   • Built the backend for a voice AI app that connected speech-to-text, intent detection, and app actions across 60+ user flows.
   • Worked on real-time calling and messaging with WebSockets and Agora SDK, adding retries and heartbeat checks so the app handled weak networks better.
   • Built an MCP-based layer for connecting agents to services, with clear events and error handling across multiple backend services.
3. Jun — Aug 2025 Zeru

   AIML Engineer Intern

   • Built a Python data pipeline for bot detection across 5 blockchain protocols, with separate preprocessing steps for each protocol.
   • Added tracking for each transaction from ingestion to final label, making it easier to debug and audit the pipeline.
   • Added validation checks and clearer logs so pipeline failures were easier to catch and fix.
4. Aug — Dec 2025 PES University

   Teaching Assistant — Computational Data Science

   Under Dr. Bhaskarjyothi Das

   • Helped students with data science and ML work, including cleaning data, building features, choosing models, and evaluating results.
   • Reviewed student projects for data leakage, weak baselines, unclear metrics, and claims that were not supported by the results.
   • Mentored teams on project structure, debugging, and algorithm implementation; also helped keep grading consistent across 250 submissions.
5. Jun — Jul 2025 CIE, PES University

   Teaching Assistant

   • Guided student teams building a blockchain-based learning platform, especially the link between smart contracts, the UI, and failure handling.
6. Aug — Nov 2024 Grain For All

   Engineer Intern

   • Built a Python ETL system that turned EPA documents with different formats into clean, structured outputs.
   • Used ReAct and AutoGen for multi-step document processing, with checks and logs at each step to catch bad outputs early.
   • Worked with different teams to update parts of the pipeline without breaking later steps.
7. Jun — Jul 2024 CDSAML, PES University

   Summer Intern

   • Built a chatbot using Ollama and LlamaIndex with a RAG pipeline, so answers could be grounded in retrieved documents.
   • Compared different RAG pipeline setups to understand which worked better for different use cases.

• 2022 — 2026

  B.Tech — Computer Science & Engineering (AI & ML)

  PES University · Bengaluru

  • DAC Scholarship recipient
  • Head of Sponsorship, Aatmatrisha '25
  • Head of Sponsorship, Chords '24
  • Head of Social Media, AIKYA