Mohit Singh

• Architected a production multi-tenant B2B SaaS platform (AI Hotel Receptionist) enabling hotels to deploy
branded AI agents for voice calls, web chat, WhatsApp, and SMS — operating 24/7 on FastAPI, async
Python, and Railway-hosted microservices.
• Engineered a real-time voice AI pipeline (PSTN → Twilio → Deepgram STT → GPT-4o tool-call loop →
ElevenLabs TTS → audio stream) achieving sub-900ms end-to-end latency — replaced LangChain with
a lean 80-line async loop to eliminate framework overhead.
• Built a full RAG pipeline (document chunking → text-embedding-3-small → ChromaDB vector store per
tenant → semantic retrieval) and a Celery async task system for booking confirmations, analytics, and
document ingestion with automatic retries.
• Delivered data analysis and customer success reporting, translating platform usage metrics into
actionable insights for client stakeholders.

• Designed and evaluated ML models across 3+ research initiatives, applying hyperparameter optimisation
and ablation studies to identify optimal configurations, with all findings documented for full reproducibility.
• Engineered LLM-based workflows using prompt engineering techniques (few-shot, chain-of-thought),
systematically improving output quality and consistency across production research pipelines.
• Built end-to-end data preprocessing pipelines spanning structured, unstructured, and time-series
modalities, reducing data preparation overhead and accelerating experimentation cycles.

• Delivered data analysis and SQL-based reporting for enterprise clients, extracting and validating
multi-source datasets to produce business-critical insights that informed strategic decision-making.
• Developed as an SAP ABAP Developer, building and maintaining custom ABAP programs, reports, and
data extraction routines to support client ERP requirements.
• Built ETL workflows and structured dashboards to communicate analytical findings to both technical and
non-technical stakeholders across cross-functional project teams.