# Anubhav Tripathi

Site: https://anubhavtripathi.vercel.app/

> Senior AI/ML Engineer at UltraSafe AI (us.inc), based in Mumbai, India, hybrid with the
> Dubai team. Fine-tunes open-source large language models into domain experts, builds
> retrieval-augmented generation systems, and builds real-time voice agents. Work shipped
> into UAE government departments, US healthcare teams and enterprise environments.
> Professional since September 2024.

## Contact

- Email: anubhavtripathi0509@gmail.com
- LinkedIn: https://www.linkedin.com/in/anubhavtripathi-ai
- GitHub: https://github.com/anubhavtripathi0509

## Current role

Senior AI/ML Engineer, UltraSafe AI, March 2026 to present. Dubai, UAE, hybrid from Mumbai.
Duplex audio and agent orchestration behind UltraVoice, on a stack where the LLM, TTS and
STT are built in-house.

## Previous roles

AI/ML Engineer, Nimap Infotech, February 2025 to February 2026, Mumbai. Client-facing
delivery of LLM systems for UAE government, US healthcare and enterprise: domain-specific
expert models, RAG retrieval, tool-augmented workflows.

Software Engineer (contract), SoftDigits, September 2024 to January 2025, remote.
GAN-based image authentication in PyTorch, deployed as a REST API.

## Projects

UltraVoice (Voice AI, 2026, UltraSafe AI, https://ultravoice.us.inc/). Voice agents that
answer real phone calls. Duplex audio path with streaming STT and TTS and barge-in,
concurrency and back-off under TTS rate limits, agent orchestration with state, tool calls,
retries and mid-call recovery.

Patient report retrieval (Healthcare, 2025, US client). RAG pipeline giving clinicians
patient records in seconds: embeddings, vector search, PostgreSQL, Django REST, with
backend validation over LLM output for access-controlled medical data.

Dubai Government service classification (Government, 2025, UAE). Fine-tuned expert model
routing service requests: SFT with PEFT and LoRA, constrained inference for deterministic
classification.

Image-based authentication (Computer vision, 2024, SoftDigits). GAN model in PyTorch behind
a REST API.

## Skills

Models and agents: SFT, PEFT/LoRA, expert models, dataset preparation, agent workflows,
prompt engineering, tool calling, LangChain, LangGraph.
Retrieval and data: embeddings, vector search, Redis, PostgreSQL, MySQL.
Voice: streaming STT, neural TTS, barge-in and endpointing, telephony transport, WebSockets.
Serving and backend: vLLM, TGI, GPU serving, AWS EC2 and S3, Docker, Python, Django REST,
FastAPI, SQL.

## Achievements

District Winner, Maharashtra Student Innovation Challenge, selected from 9,227 teams.
Winner, Technoscope 2024, among 800+ participants.

## Education

B.E. Computer Engineering, Anjuman-I-Islam's Kalsekar Technical Campus, Navi Mumbai,
2021 to 2025, CGPA 7.52 of 10.