education
msc advanced computer science (ai)
university of leeds | 2025 — 2026
b.tech ai & data science
viit, pune | 2021 — 2025
experience
founding ai engineer · hopps
september 2024 — june 2025 | pune, india
- led cross-functional product development for a food-tech platform serving ::y::70+ restaurant partners:: and 20+ beta users.
- engineered an ai-powered recommendation system using ::b::gemini api::, firebase, and flutter to deliver personalized dining suggestions.
- architected evolution from keyword extraction to ::p::embedding-based semantic retrieval::, identifying production bottlenecks in api latency and cost-per-query constraints.
- implemented production safeguards including ::y::rate limiting:: (3 ai bio generations/user/day) and api key rotation for cost management during beta.
- mentored 2 engineering interns on recommendation system architecture and mobile development, conducting code reviews and technical guidance.
research intern · iim udaipur
september 2023 — april 2024 | remote
- built an automated classification system processing ::y::250,000+ nft records:: using python and hugging face to differentiate ai-generated vs. human-created content.
- engineered asynchronous data collection infrastructure using ::b::asyncio and multithreading:: to scrape 113,000+ web pages with optimized throughput.
- explored cloud-based distributed scraping architecture on ::p::gcp cloud functions:: for ip rotation and scalable data collection.
- developed automated preprocessing workflows applying nlp techniques that led to ::y::two research manuscripts::.
ai intern · konverge.ai
august 2023 — october 2023 | nagpur, india
- applied statistical modeling and machine learning for customer segmentation, purchase prediction, and profitability analysis.
- built ::b::sentiment classification model:: using nlp techniques to analyze user reviews and extract actionable insights.
- developed data preprocessing pipelines and visualizations to communicate model performance and business recommendations.
projects
maira · ai companion system live demo
- engineered a ::p::rag-based memory system:: integrating ::b::gemini 2.0 api:: with sqlite embedding retrieval using semantic similarity thresholds for context injection.
- designed a parallel memory pipeline leveraging gemini to analyze message history, extracting ::y::categorized memories with importance scoring:: and redundancy detection.
- built full-stack architecture using flutter, firebase, getx, and gcp cloud functions to track persistent user context and emotional states.
- implemented fluid chat ux inspired by ::p::imessage design principles:: for judgment-free companionship with long-term context persistence.
social media recommendation engine blog live demo
- engineered a tiktok-style recommendation engine using ::y::python, flask, and faiss:: for semantic video retrieval.
- implemented ::b::indexivf clustering:: with approximate nearest neighbor search to efficiently query 3072-dimensional embeddings across 500+ videos.
- designed ::p::sliding window user preference model:: (last 5 interactions) with 70-30 exploitation-exploration split to balance relevance and discovery.
- built multi-signal scoring system across 6 interaction types to capture nuanced user preferences.
brewstories · llm fine-tuning for creative text generation
- fine-tuned ::y::tinyllama transformer (1.1b parameters):: on 10,000-story dataset using pytorch and hugging face transformers.
- developed end-to-end training workflow including dataset preparation, tokenization, fine-tuning, and evaluation for generative text modeling.
- achieved ::b::runner-up recognition:: at visionary techfest 24 ai/ml hackathon organized by binghamton university.
distributed training infrastructure · fraud detection full project overview
- architected ::p::federated learning system:: enabling privacy-preserving collaborative training across 3 distributed clients over 10 rounds.
- engineered data processing pipeline for highly imbalanced ethereum datasets (90%+ class imbalance), applying ::y::smote oversampling:: and feature engineering.
- developed fraud detection classifier achieving ::b::99.7% auc-roc and 96.6% recall:: on transaction pattern analysis.
- built flask-based visualization dashboard to monitor performance metrics and communicate fraud pattern insights.
research
“attention span and its correlation with mental health”
richa shah, sarthak rak, ayush patne, dr. laxmi bewoor
-
4th asian conference on innovation in technology, august 2024
- led research design analyzing correlation between attention span metrics and mental health indicators across 84 participants.
- defined analytical approach utilizing ::p::knn clustering:: for behavioral segmentation and pattern identification.
technical skills
- programming: python (pandas, numpy, scikit-learn), pytorch, tensorflow, hugging face transformers
- ai/ml: nlp, generative ai, llms, transformer architectures, prompt engineering, fine-tuning (lora/peft)
- infrastructure: aws, gcp (cloud functions, firebase), langchain, flask, git, distributed training systems, model deployment
- specializations: rag systems, semantic similarity, asynchronous data pipelines, model evaluation
leadership & achievements
- hackathon lead organizer — viz-a-thon (dataviz), viit
- runner-up — ai/ml domain, visionary techfest 24 hackathon (binghamton university)
- media team head — tedxviit
- video & photography head — ai student association, viit
- class representative — ay 2021-22
certifications
- machine learning specialization — stanford online & deeplearning.ai
- generative ai with large language models — deeplearning.ai & amazon web services