I write code, learn systems, and ship projects while preparing to move from classroom work to professional engineering.
Projects reflect applied software work from my classes and personal efforts, focused on coding, systems, and problem solving as I gain practical experience.
Designed an NLP evaluation experiment probing whether GPT-4o judges multi-turn conversation quality holistically or locally. Constructed 8 multi-turn dialogues across 4 domains (math, coding, factual, emotional) with detailed ground-truth quality labels. Measured a 0.33-point quality delta after turn shuffling, exposing GPT-4os local-quality bias as an automated evaluation judge.
Architected an AI recommendation platform with Gemini LLM and Spoonacular API, serving 50+ users personalized weekly meal plans. Tuned prompt engineering and constraint-based filtering, cutting irrelevant recommendations 35% across diverse dietary user queries. Deployed to Streamlit Community Cloud with API keys secured via environment variables, exposing zero credentials publicly.
I write code, learn systems, and ship projects while preparing to move from classroom work to professional engineering.
I am Anushka Jain, a third-year Computer Science student at the Georgia Institute of Technology. I have some hands-on experience and I am building toward a professional career in software and systems, exploring projects that connect code, design, and real problems.
Experience
AI/ML Engineering Intern
North Star AI Labs, ATDC · May 2026 Present
iOS Developer, Yes Chef
iOS Club at Georgia Tech · Oct. 2025 Dec. 2025
Full-Stack Developer, Pin
WebDev at Georgia Tech · Sep. 2025 Oct. 2025
Software Engineering Intern
Aadhar Housing Finance Ltd. · Jun. 2025 Jul. 2025
Education
Georgia Institute of Technology
Bachelor of Science in Computer Science, Honors | Threads: Intelligence & Information Internetworks | GPA: 4.00 · Class of 2028
Skills
Designed an NLP evaluation experiment probing whether GPT-4o judges multi-turn conversation quality holistically or locally. Constructed 8 multi-turn dialogues across 4 domains (math, coding, factual, emotional) with detailed ground-truth quality labels. Measured a 0.33-point quality delta after turn shuffling, exposing GPT-4os local-quality bias as an automated evaluation judge.
Architected an AI recommendation platform with Gemini LLM and Spoonacular API, serving 50+ users personalized weekly meal plans. Tuned prompt engineering and constraint-based filtering, cutting irrelevant recommendations 35% across diverse dietary user queries. Deployed to Streamlit Community Cloud with API keys secured via environment variables, exposing zero credentials publicly.
Engineered a computer-vision Discord bot detecting dermatological markers via image analysis, achieving 80% detection accuracy. Integrated Google Maps geolocation APIs to surface the 3 nearest specialized medical facilities per query in under 2s. Configured an OpenCV pipeline using color-space conversion to isolate lesion regions, stabilizing detection across diverse inputs.