Raj Shah

Python Developer and AI/ML Engineer with production experience building voice AI platforms for enterprise telephony — covering call routing, real-time LLM/ASR/TTS pipelines, and FreeSWITCH-based dialer integrations. Designed and deployed a containerized voicebot with multi-provider LLM orchestration (Gemini 2.0 Flash, Groq LLaMA 3.3 70B, Ollama), RAG-based knowledge retrieval via Redis, barge-in handling, and real-time noise cancellation. Proficient in Python, Flask, FastAPI, Django, and PostgreSQL across the full stack.

Comfortable using AI coding tools (Claude, Cursor) to build faster. Eager to ship real features, grow across the stack, and build intelligent systems real customers use.

Jr. Developer - Unified Technology Pvt. Ltd.

(2025-10)

• Designed and deployed a production-grade AI voicebot platform for CPaaS — handling call routing, SIP session management, and LLM-driven conversational flows for enterprise clients, replacing legacy IVR systems with conversational AI.
• Multi-provider LLM orchestration: Configured Gemini 2.0 Flash as primary, Groq LLaMA 3.3 70B as automatic fallback, and Ollama for offline operation; implemented background health checks and seamless failover ensuring uninterrupted call processing.
• RAG-based knowledge retrieval: Built Redis-backed per-call conversation history retrieval with domain knowledge-base injection into LLM prompts, enabling context-aware, grounded responses and eliminating hallucination on known product/service topics.
• Implemented FreeSWITCH dialplan configuration, dialer logic, and campaign-ready call routing; built custom mod_audio_fork module to stream live call audio to AI pipelines with low-latency real-time processing.
• Integrated Silero VAD (per-call state isolation) and DeepFilterNet2 real-time noise cancellation on raw telephony audio, significantly improving STT accuracy across concurrent calls in noisy environments.
• Built barge-in interruption handling enabling natural mid-response caller interaction; debugged production audio pipeline failures by tracing WebSocket logs, STT, LLM, and TTS layers end-to-end.
• Maintained call detail records (CDRs) and structured session logs for operational debugging, reporting, and analytics on live call traffic.
• Containerized the full voicebot platform with Docker; published images to Docker Hub; managed deployment via Systemd services and Cron scheduling.

Python Intern - Unified Reach Pvt. Ltd.

(2025-03 - 2025-09)

• Developed a Python-based AML (Anti-Money Laundering) verification agent for onboarding fraud detection, integrating watchlist matching and entity resolution.
• Implemented cosine similarity scoring and rule-based risk checks for NLP-driven classification of high-risk entities; added structured logging for audit trail and operational debugging.
• Conducted NLP model research and training for text classification and intent-detection tasks; validated agent decision logic with pytest-based unit tests.

B.Tech - Computer Engineering - Lok Jagruti Institute of Engineering and Technology (2021-11 - 2025-06)

HSC (Science) - Science - Kendriya Vidyalaya Sangathan (2019 - 2021)