AI Engineer with less than a year in Audio QA and RAG systems.
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Sarthak Senapati is an aspiring Software Engineer with 3 months of internship experience at MediaTek, specializing in acoustic framework development and integration into chipset validation pipelines. He has also developed advanced projects including an Agentic RAG System with multilingual support and a Multi-Modal Biometric Tracking System, demonstrating proficiency in AI/ML, full-stack development, and cloud deployment. Currently pursuing a Bachelor of Technology in Computer Science, Sarthak is skilled in Python, LLMs, and computer vision, focusing on building production-grade intelligent systems.
Kalinga Institute of Industrial Technology
Bachelor of Technology · Computer Science
September 1, 2022 – May 1, 2026
MediaTek
Software Engineer Intern
May 1, 2025 – August 1, 2025
Noida, Uttar Pradesh, India
Agentic RAG System with Multilingual Support
May 1, 2026 – Present
Built production-grade RAG over a 476-page corpus achieving faithfulness 0.940 and answer relevancy 0.785 via async RAGAS evaluation pipeline with Groq llama-3.3-70b as judge LLM. Implemented LangGraph agent with Groq-powered LLM query classifier routes simple queries to direct retrieval and complex queries to full hybrid dense + BM25 sparse search with cross-encoder reranking over top-20 candidates. Integrated Sarvam Translate API enabling queries across Indian languages Hindi input translated to English for retrieval, answers returned in source language with toggleable language mode. Deployed FastAPI backend on Railway with Qdrant Cloud for vector storage, PostgreSQL for eval logging, CI/CD via GitHub Actions, streaming responses with zero latency impact. Architected LangChain + LlamaIndex ingestion pipeline with semantic chunking across 1365 chunks and Streamlit evaluation dashboard showing per-query faithfulness and relevancy scores.
View ProjectMulti-Modal Biometric Tracking System
March 1, 2026 – Present
Built a real-time multi-modal biometric identification system fusing face recognition (InsightFace), body ReID (OSNet-x1.0), and gait analysis, achieving >90% identification accuracy at <5% false-positive rate and 15–20 FPS on GPU across multi-camera environments. Engineered a YOLOv8 + ByteTrack detection pipeline with a weighted fusion engine using cosine similarity over 512D/2048D embeddings; diagnosed and fixed a partial OSNet weight-loading bug (522-567 keys) that recalibrated optimal modality weights from Face 95%/Body 4% to Face 65%/Body 35%, substantially improving body-only identification accuracy. Architected a production backend (FastAPI, PostgreSQL, Redis) orchestrated via Docker Compose across tracker, backend, and Streamlit dashboard services; deployed with CI/CD on GitHub Actions and instrumented a Prometheus + Grafana observability stack tracking per-camera FPS, matcher latency, and camera health.
View ProjectCultural Fit Analysis
The candidate's personal projects showcase a strong initiative and passion for AI engineering, aligning well with a role that requires self-driven exploration and implementation. The diversity of projects (RAG, biometrics, audio processing) indicates a broad interest and adaptability. Their experience with collaborative tools like GitHub Actions and Docker suggests readiness for team environments and modern development workflows.
Soft Skills & Operational Fit
The candidate demonstrates strong problem-solving skills through debugging complex issues and designing robust systems. Their project descriptions indicate a proactive approach to building production-grade solutions and a good understanding of operational aspects like observability and deployment. The MediaTek internship shows an ability to integrate into existing production pipelines and deliver measurable impact.