Help & Supportcareers@gravityer.com

Siranjeevi P | Gravity Talent Portal

SP

Siranjeevi P

AI Engineer

https://talent.gravityer.com/siranjeevi-p

AI Engineer with less than a year in LLMs & RAG Pipelines

F22 Labs

AI is analyzing your overall score…

Identifying your key strengths…

Evaluating your skill match against the job requirements…

Assessing your cultural and operational fit

Member since June 24, 2026

https://talent.gravityer.com/siranjeevi-p

About

AI Engineer Intern with experience building LLM-powered applications, retrieval systems, and full-stack web solutions. Worked on embedding model fine-tuning, RAG pipelines, prompt evaluation systems, and AI-driven developer tools. Skilled in Python, Java, JavaScript, React.js, Node.js, and Spring Boot, with a focus on building practical and scalable software.

Top Skills

PythonJavaJavaScriptReact.jsNode.Js

Skills

PythonJavaJavaScriptReact.jsNode.JsSpring BootMysqlMongoDBStreamlitNltkExpress.Jsspring-securityjwt

Education

Anand Institute of Higher Technology

B.E. · Computer Science Engineering

August 1, 2022 – June 30, 2026

Experience

F22 Labs

AI Engineer Intern

December 1, 2025 – June 30, 2026

India

VCodez

Machine Learning Intern

August 1, 2025 – October 31, 2025

India

Projects

Embedding Model Fine-Tuning for JD-Resume Matching

December 1, 2025 – June 30, 2026

Fine-tuned Qwen3-Embedding-0.6B using two-phase LoRA training (~4.5M params) with MultipleNegativesRankingLoss on JD-resume triplets to improve semantic role separation in tech recruitment retrieval. Engineered a synthetic data pipeline using Groq LLM to generate role-specific JDs, map domain-confusor hard negatives, and rebalance 20+ role categories to prevent majority-domain overfitting; dataset published to Hugging Face. Phase 2 model scored 101/125 on held-out RAG eval vs. 86/125 for base, eliminating catastrophic forgetting from Phase 1. Benchmarked against 5 open-source models across 394 pairs; achieved 98.73% accuracy and separation score of 28.97 vs. 16.27 base.

Prompt Evaluator, Fixer & Runner

December 1, 2025 – June 30, 2026

Built a Prompt Evaluator that compares call transcripts against system prompt requirements and reports instruction violations in a structured format. Developed a Prompt Fixer module that generates revised prompt outputs for detected violations and supports streaming inference through multiple LLM providers (mimo-v2.5 and Gemini 2.5 Flash) with token usage tracking. Engineered a Prompt Runner via a Node.js backend that executes configured prompt templates against LLM endpoints and stores results in a database layer, backed by a JS/CSS frontend for interactive review.

Skill Connect - Skill Exchange Platform

December 1, 2025 – June 30, 2026

Built a full-stack skill exchange platform with a Spring Boot backend and React.js frontend for video-based skill discovery and profile management. Implemented JWT-based authentication with Spring Security: signed token issuance on login, per-route validation, and role-based access control. Developed RESTful APIs for video upload, skill categorization, and keyword search backed by MongoDB; implemented protected routes and token-based session handling in the React frontend.

Slack RAG Bot

December 1, 2025 – June 30, 2026

Built a Slack-integrated RAG bot using LangChain and FAISS to retrieve relevant context from internal knowledge bases and deliver grounded context-aware responses in-channel. Implemented full document ingestion pipeline: chunking, embedding, and vector storage for fast semantic search at query time. Designed thread-aware reply logic ensuring responses post back to the originating Slack thread, keeping channel context clean and traceable.

SmartQ - Voice-Enabled Queue Management System

December 1, 2025 – June 30, 2026

Built a full-stack queue management platform with a Node.js/Express REST API, React.js frontend, and a Python voice agent powered by LiveKit Agents. Integrated Groq LLM (Llama-3.3-70b-versatile), AssemblyAI STT, Silero VAD, and Cartesia TTS for a real-time voice interaction pipeline with per-session latency logging. Supported QR-based queue joining, priority levels (normal/urgent/VIP), and shop/room management with noise cancellation for real-time interactions.

Certifications

F22 Labs AI Engineer Intern Certificate

F22 Labs

January 1, 2026 – Present

VCodez ML Internship Certificate

VCodez

January 1, 2025 – Present

Key Strengths

Extensive experience in LLM-powered application development, including fine-tuning embedding models and building RAG pipelines.
Proficient in prompt engineering, evaluation, and fixing, demonstrating a deep understanding of LLM interaction and optimization.
Strong full-stack development skills (Node.js, Spring Boot, React.js) combined with AI expertise, enabling end-to-end system delivery.
Demonstrated ability to integrate various AI services (STT, TTS, VAD, LLMs) into real-time interactive systems.
Experience with synthetic data generation and dataset rebalancing for robust model training, indicating a strong grasp of ML data pipelines.

Cultural & Operational Fit

Cultural Fit Analysis

The candidate's diverse project portfolio, ranging from embedding model fine-tuning to full-stack queue management systems and RAG bots, demonstrates adaptability and a broad interest in various technical domains. The focus on practical, scalable solutions aligns well with a product-driven environment. The candidate's involvement in publishing blogs also indicates a willingness to share knowledge and contribute to the technical community.

Soft Skills & Operational Fit

The candidate's project descriptions highlight a proactive approach to problem-solving and a strong ability to deliver complex technical solutions. Participation in hackathons and competitive programming suggests a collaborative and results-oriented mindset. The detailed project descriptions indicate good communication of technical concepts.