Help & Supportcareers@gravityer.com
AI Engineer with less than a year in LLMs, RAG & LangChain
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Highly motivated AI/ML Developer Intern with a focus on Large Language Models (LLMs), Prompt Engineering, and Retrieval-Augmented Generation (RAG). Experienced in building multi-agent conversational systems using LangGraph, LangChain, and Groq, and developing self-corrective RAG pipelines. Possesses strong skills in Python, SQL, and various ML frameworks, with a proven ability to deliver production-grade AI solutions for complex business challenges.
GUVI HCL IITM
AI and Machine Learning Program · Artificial Intelligence and Machine Learning
August 1, 2024 – June 30, 2025
Rajalakshmi Engineering College
Bachelor of Engineering (B.E) · Computer Science and Engineering
August 1, 2021 – June 30, 2024
Least Action Company
AI/ML Developer Intern
July 1, 2025 – October 1, 2025
India
RAG Tech Assistant - Self-Corrective Technical Q&A
June 1, 2026 – Present
Built a self-corrective RAG pipeline using LangGraph - 6 nodes covering query analysis, retrieval, per-chunk LLM grading, answer generation with inline citations, hallucination checking, and a smart retry loop that rewrites queries using grade reasoning from failed attempts before falling back to Serper web search. Developed a FastAPI backend with 9 production-grade endpoints for document ingestion (URLs, PDFs, .docx, .md), session management, and feedback collection; implemented two-layer duplicate detection (SHA-256 document checksums + deterministic chunk IDs in Chroma Cloud) to minimize re-embedding costs. Uses Cohere embed-v4.0 for semantic embeddings in Chroma Cloud, PostgreSQL for persistent chat history outside the graph, and a Streamlit UI exposing per-chunk grade reasoning - making the self-corrective pipeline observable and debuggable.
View ProjectDex - Multi-Agent AI Executive Assistant
January 1, 2026 – Present
Built a multi-agent, LLM-powered AI Executive Assistant that replaces 9 separate tools with one conversation — specialized agents handle research, drafting, planning, and image generation, with user confirmation for sensitive actions. Powered by LangGraph with a Supervisor-Worker-Automation architecture, using prompt engineering to define agent roles and tool-calling behavior; Worker handles user tasks and hands off to Automation, which executes actions silently in the background. Integrates Google OAuth and Gmail APIs with separate login and Gmail connect flows; secures user data with Fernet encryption for tokens and emails at rest and SHA256 email hashing. Deployed on Render with PostgreSQL on Neon; LangSmith provides tracing and evaluation for every agent decision, with per-user Groq rate-limit fallback for reliability under load.
FlipKartCSAT — AI-Powered Customer Satisfaction Intelligence
May 1, 2025 – May 1, 2025
Helps managers identify dissatisfied customers before escalation — built an XGBoost classification model to predict CSAT from unstructured support text with 94% accuracy, after preprocessing and handling data with a custom Column Transformer (TF-IDF, OneHot/Ordinal encoding, StandardScaler). Cuts response time - integrated the Gemini GenAI API to auto-generate personalized customer replies and send them via SMTP email automation, eliminating manual drafting after every support case. Gives managers visibility into team performance - internal Gemini-powered (LLM) Streamlit chatbot answers agent/supervisor queries; Power BI dashboard tracks CSAT trends across shifts and teams.
View ProjectContextZoom - Transformer-Based Feedback Validation
March 1, 2025 – March 1, 2025
Validates whether a written reason actually matches its source text — fine-tuned a BERT (bert-base-uncased) transformer model for text classification, distinguishing valid vs. invalid justifications and catching mismatched explanations automatically. Handled limited labeled data with synonym replacement (30% probability), negative sampling, downsampling, and class-weighted loss; tested and evaluated model performance, achieving 91.22% validation accuracy with custom threshold tuning. Deployed as a real-time validation app on Hugging Face Spaces using Streamlit, built with TensorFlow, HuggingFace Transformers, and Adam optimizer at Ir=1e-5.
View ProjectPython Programming
Udemy
June 1, 2026 – Present
GUVI HCL IITM – AI and ML Program
GUVI HCL IITM
June 1, 2026 – Present
SaWiT.AI Challenge
SaWiT.AI
June 1, 2026 – Present
Generative AI
GUVI HCL
June 1, 2026 – Present
Cultural Fit Analysis
The candidate demonstrates a strong cultural fit for an AI Engineer role, particularly in an innovative and fast-paced environment. Their portfolio showcases a diverse range of personal projects, indicating self-motivation, initiative, and a passion for applying AI to real-world problems. The projects cover various aspects of AI, from traditional ML to advanced generative AI and multi-agent systems, reflecting a broad skill set and adaptability. The candidate's proactive learning through certifications and an ongoing AI/ML program further emphasizes their commitment to continuous growth and staying current with industry trends. The alignment of their projects with the target role's requirements is excellent.
Soft Skills & Operational Fit
The candidate's project descriptions indicate a strong problem-solving aptitude, evidenced by their ability to identify business needs (CSAT prediction, executive assistant automation) and develop comprehensive AI solutions. Their detailed explanations of handling limited data, implementing self-corrective mechanisms, and ensuring data security suggest a meticulous and robust approach to development. The use of tools like LangSmith for tracing and Streamlit for observability points to a proactive stance on debugging and operational reliability. The candidate's focus on user experience (single AI interface, personalized replies) also highlights a user-centric mindset.