Hey there, this is Ishika

Budding AI Developer
Artificial intelligence, Natural Language Processing Grad @ UCSC
From Data to Decisions to Discovery

Ishika's Profile Picture

About Me

As a grad student in NLP at UC Santa Cruz, I am passionate about building AI that is impactful. With a deep love for AI, LLMs, and machine learning, I have completed my Bachelors in technology, Computer Science Engineering with a specialization in Artificial intelligence and Machine learning, and worked as a Software Development Engineer at Signify Innovation Labs (Philips Lighting).

Beyond the code, I have led student clubs like Google Developers Student Club and Artificial Intelligence Club, organized tech events, and thrived in collaborative, high-energy environments.

Driven by curiosity and caffeine, I love solving complex problems and bringing creative ideas to life.

NLP Machine Learning Deep Learning LLMs Reinforcement Learning Data Science and Analysis Agentic Workflows

Work Experience

My professional journey so far

Software Development Engineer

Jul 2023 – Aug 2024

Signify Innovation Labs – Bangalore, India

  • Developed 3+ Android apps for intelligent workspaces, deployed to 10,000+ enterprise users for smart office operations like Booking Meeting rooms, navigation and controlling lights.
  • Strengthened code quality and reliability with 90%+ unit test coverage and integrated automated testing using JUnit and Firebase Crashlytics.
  • Accelerated delivery pipelines by 40% by designing CI/CD workflows with Jenkins, reducing deployment friction across teams.
  • Modernized legacy applications by adopting scalable MVVM architecture, enhancing code modularity and long-term maintainability.
  • Explored AI-driven features with Gemini, introducing contextual NLP enhancements that boosted app usability and user engagement.
Android Development MVVM Firebase LLMs

Research and Development Intern

Jan 2023 – Jul 2023

Signify Innovation Labs – Bangalore, India

  • Built an Android SDK powering 5+ production apps, consolidating 30+ REST APIs to accelerate development and reduce redundancy in code.
  • Integrated Zigbee device support, improving interoperability and enhancing cross-device communication efficiency by 25%.
  • Optimized SDK architecture to reduce API response time enhancing app responsiveness and user satisfaction.
REST APIs Android SDK Clean Architecture

Google Developer Student Club Lead

Aug 2020 – Jul 2022

VIT Bhopal University – Bhopal, India

  • Directed a team of 60+, organizing 15+ hands-on events including Android Study Jams and ML Talks, fostering a vibrant developers community.
  • Worked with 1,800+ students through engaging workshops, hackathons, and live coding challenges to spark interest in emerging technologies.
  • Elevated club visibility and participation through collaborative leadership and consistent engagement initiatives.
Leadership Community Building Tech Events Team Management

AI Club Co-Founder and AI Department Student Head

2020 – 2021

VIT Bhopal University – Bhopal, India

  • Co-founded the first AI-focused student community, successfully launching the AI Conclave with 500+ attendees from across the country.
  • Mentored 20+ students through weekly sessions and projects, building foundational ML skills and encouraging innovation.
  • Created blogs and educational content that tripled online reach, establishing the club as a go-to resource for AI learning on campus.
AI Mentorship Public Speaking Community Outreach

Featured Projects

A selection of my recent work

Structured Function Calling with Gemma-2B

Designed tag-driven output format and fine-tuned Gemma-2B to generate auditable function calls with reasoning traces and schema-based tool use.

Tool Use Function Calling Reasoning
View→

Quantized Llama-3.2 (8-bit weights + 16-bit activations)

Replaced linear layers in Llama-3.2 with W8A16 quantized modules for lower memory usage and faster inference with minimal accuracy loss.

Quantization Memory Optimization Layer Replacement
View→

LLM Pushback Evaluation Framework

Built experiment runner to test how LLMs respond to pushback across multi-turn chats. Uses judge models for scoring resilience and conversational integrity.

LLM as Judge LLM Pushback Conversational Assistants
View→

LoRAfy: Fine-Tuned Llama-3.2

Applied LoRA to fine-tune Llama-3.2-1B for improved function calling and structured external tool interaction.

LoRA Retrieval Function Calling
View→

Shakespeare GPT-2 (From Scratch)

Built and fine-tuned GPT-2 from scratch on Shakespeare's works with custom tokenization, dataset binning, and character-level training.

GPT-2 Fine-Tuning Paper Implementation
View→

CharacterVerse — On-Device LLMs

Android app running LLMs locally with character-specific system prompts for dynamic, persona-based conversations.

Android Kotlin On-Device LLM
View→

SemEval 2024 Task 7 — Fact-Check IR

Built multilingual classifiers for retrieving relevant fact-checks for social media posts across 14 languages using SemEval 2025 dataset.

Fact-Checking Multilingual NLP ROBERTa
View→

Get In Touch

I'd love to hear from you

Email

ikulkar1@ucsc.edu

Phone

+1 (408) 581-2315

Location

San Francisco Bay Area, CA