Retrieval QA with LLaMA2-7B
Summary
Augmenting LLM with context to increasing accuracy of answer
About
A Computer Science Engineering student with experience in AI, full-stack development, and financial data analysis.
Work
Whitetable.ai
|Founding Engineer
Summary
Led the development of a hiring intelligence incoporating latest research in multi-agent AI systems, with streamlined deployment and iteration using Jenkins and Docker. Designed an adaptable orchestration system, dynamically overseeing a multitude of 20+ diverse agents in real-time, and optimizing 100s interview scenarios with a staggering 90% automation proficiency. Enhanced generation accuracy and reduced hallucination errors by 30% through the integration of ReAct agents and advanced retriever tools, surpassing the capabilities of conventional LLM chains.
Highlights
Fine-tuned the Gemini Flash model for resume parsing, improving the model's ability to extract and structure candidate information for job-matching scenarios.
Delivering a Critical AI System Under Tight Timelines
Built a custom ReAct agent with controlled prompting and parameter tuning to stabilize the results.
Integrated KSA (Knowledge, Skills, Abilities) framework, and integrated it into algorithm, ensuring it accurately reflected the true candidate-job fit.
The system was successfully launched on time, now screening 100s of candidates daily. It has become one of Whitetable.ai's most impactful tool, significantly improving the hiring process.
WhizIT
|Full Stack Developer (winter intern)
Summary
Implemented complex and modular form handling, leveraging both the Tanstack Table for advanced data display and React Hook Form for efficient data management. This approach ensures a scalable and maintainable codebase.
Highlights
Integrating Server-Side Rendering (SSR) led to a reduction in initial page loading time.
Metadata and SSR allow search engines to easily crawl and index pages, positively impacting organic search traffic.
Designed such that Next.js significantly enhanced SEO by efficiently indexing more pages
Nuvama (formerly Edelweiss)
|Summer Analyst
Summary
Mastered financial market concepts and quantitative topics, showcasing proficiency in handling data.
Highlights
Transformed data from various sources into line protocol streamlining the analysis of 50,000+ records.
Used InfluxDB for efficient data management, working on streaming and static data with sub-second accuracy.
Produced impactful visualizations that communicated patterns and trends from complex datasets.
Education
India Institute of Information Technology, Pune
→
B-Tech
Computer Science Engineering
Grade: 7.9
Skills
Developer Tools
Github, VS Code, VIM, Docker, Docker-volumes, UNIX, Jenkin.
Programming Languages
Python, C++, SQL, JavaScript, HTML5, CSS, Bash.
Frameworks/Libraries
Langchain, Langgraph, Langfuse, Crew AI, Express, React, Node, sk-learn, NumPy, Pandas, Hardhat.
Databases
MongoDB, MySQL, InfluxDB, Pinecone.
Models
Gemini-flash, Claude, LSTM, FBprophet, LightGBM, GaussianNB, XGBoost, TabNet.
Coursework
Maths, Java, Data Base Management System, Object Oriented Programing, SDLC, TOC, Operating system.
Soft Skills
problem-solving, communication, teamwork, analytical skills.