Help & Supportcareers@gravityer.com
Full Stack Engineer with less than a year in LangGraph & Microservices.
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Final year B.Tech student in AI and Data Science with 8.72 CGPA from VIIT Pune. Architected and deployed 3 full-stack AI systems spanning LangGraph orchestration, RAG with Qdrant vector search, multilingual ASR pipelines (Deepgram Nova-2) with PII redaction, and real-time, event-driven backends (Socket.IO and SSE). Built a GPU-accelerated 3D portfolio with React Three Fiber achieving 60 FPS rendering. Proficient in Prompt Engineering (structured output, Pydantic guardrails), Docker containerization, Kubernetes orchestration, and distributed Microservices deployed across Render and Vercel.
Vishwakarma Institute of Information Technology (VIIT), Pune | Savitribai Phule Pune University
Bachelor of Technology · Artificial Intelligence and Data Science
August 1, 2023 – May 1, 2027
Jawahar Navodaya Vidyalaya, Krishna
Higher Secondary Certificate (HSC) · Science
July 1, 2022 – May 1, 2023
BESRC (ISRO affiliated)
DEF-Space Winter Intern
December 1, 2025 – January 1, 2026
India
BlockseBlock Technologies
AI and Decentralized Systems Intern
August 1, 2025 – September 1, 2025
India
TalentAI - Semantic Recruitment Intelligence Platform
June 24, 2026 – Present
- Orchestrated a 4-agent LangGraph StateGraph workflow orchestration (Extractor, Scorer, Fraud Detector, Ranker) processing 768-dimensional Gemini embeddings with conditional error routing to a failure-handler node (directed acyclic graph); deployed as a distributed 3-tier architecture across Vercel, Render Node.js Gateway, and Render Python AI Service. - Developed a RAG pipeline using Qdrant vector database with Gemini embeddings and HNSW-based ANN search with cross-encoder reranking for semantic candidate retrieval over 512-token overlapping chunks; enforced structured LLM outputs via Pydantic schemas with zero-shot prompt guardrails. - Designed a polyglot persistence (multi-database) layer (PostgreSQL via Prisma, MongoDB via Mongoose, Qdrant, ChromaDB fallback) with async BullMQ and Redis job queues eliminating HTTP 504 timeouts on LLM heavy operations; validated across 17 E2E workflows achieving 100% pass rate. - Instrumented with Prometheus metrics and 5 Grafana dashboards tracking queue depth, AI latency, and billing; authored 7 k6 load-test scripts targeting 1,000 concurrent users with under 200-millisecond search SLA; configured Kubernetes HPA scaling from 3 to 15 pods at 75% CPU threshold with fault-injection scripts simulating pod kills, 200ms network-latency injection, and memory pressure to validate graceful degradation.
View ProjectAI Call Analytics - Call Center Quality Auditor
June 24, 2026 – Present
- Automated 100% of call quality assurance (replacing manual sampling) with a Deepgram Nova-2 ASR pipeline featuring speaker separation across 3 languages, PCI, SSN, and numeric PII redaction at API boundary, and regex-based phonetic normalization, achieving 94% Word Error Rate (WER) accuracy and reducing transcription errors by 40% for Hindi and Tamil business terms. - Engineered a multilingual ASR pipeline supporting Hindi, Tamil, and English with custom phonetic normalization, word-level deduplication, and schema-enforced LLM intent classification covering 4 payment types and 5 rejection categories for deterministic compliance scoring. - Achieved sub-second API responses by offloading Deepgram and Gemini inference to FastAPI Background Tasks; built a deterministic regex fallback engine ensuring graceful degradation when the LLM is unavailable; Dockerized with a Python 3.11 slim image including ffmpeg and Tesseract OCR; deployed live on Render.
View ProjectEmergency Response System - Real Time Command Center
June 24, 2026 – Present
- Architected a dual-channel real-time system using Socket.IO WebSockets and SSE fallback with JWT authenticated handshake, broadcasting an event-driven architecture covering 20+ event types to multiple agency dashboards; deployed an event replay buffer (event sourcing) storing 1,000 events over a 60-minute window, persisted in Redis, for client state synchronization on reconnect. - Configured PostGIS geospatial queries for nearest hospital matching and green-corridor traffic-signal preemption simulation; applied Redis caching with a 30-second TTL on paginated list endpoints, using filter-aware cache keys and mutation-triggered invalidation for zero-stale-read guarantees. - Integrated Twilio multi-channel alerts covering voice, SMS, and email; built NLP-based AI triage endpoints for incident classification and weather-severity scoring; generated Swagger and OpenAPI documentation with full audit logging tracking actor ID, role, IP, and request ID across an event-sourced (CQRS-style) mutation log covering 25+ incident and vehicle operations.
View ProjectDeep Learning Specialization (5 Courses)
DeepLearning.AI (Andrew Ng), Coursera
June 1, 2026 – Present
Data Science, ML, NLP and Deep Learning Bootcamp
Udemy (Krish Naik)
June 1, 2026 – Present
Full Stack Development
Unstop
June 1, 2026 – Present
Cultural Fit Analysis
The candidate's diverse project portfolio, ranging from semantic recruitment to call analytics and emergency response systems, demonstrates adaptability and a broad interest in applying AI/ML to various domains. The involvement in internships with ISRO-affiliated organizations and BlockseBlock Technologies, along with participation in hackathons and open-source contributions, indicates a strong drive for continuous learning and a collaborative spirit. The target role of 'AI/ML Engineer (Graduate)' aligns perfectly with the candidate's academic background in AI and Data Science and their practical experience.
Soft Skills & Operational Fit
The candidate's project descriptions indicate strong problem-solving skills, a proactive approach to system design challenges (e.g., eliminating HTTP 504 timeouts, graceful degradation), and a commitment to robust testing and monitoring. The detailed explanations of complex architectures suggest good communication of technical concepts. The involvement in hackathons and open-source contributions points to a collaborative and innovative mindset.