Machine Learning Engineer

Abhishek Vishwakarma

I specialize in

Specialized in building intelligent conversational AI systems, RAG applications, and voice AI platforms using LLMs, LangChain/LangGraph, and modern ML frameworks.

About

A bit about me

I'm a Jr. Machine Learning Engineer specializing in developing production-ready AI systems, fine-tuning LLMs, building RAG applications, and creating voice AI platforms.

Currently working at Anvex AI Technologies, I architect end-to-end ML pipelines, integrate cutting-edge AI models, and build scalable backend systems for conversational AI and voice applications. My work spans from RAG-powered chatbots to real-time voice AI systems with custom TTS models.

With a B.E. in AI & Data Science (9.20 CGPA, Department Rank 1), I combine strong academic foundations with hands-on experience in PyTorch, Transformers, LangChain/LangGraph, and modern MLOps practices to deliver innovative AI solutions.

LocationMumbai, Maharashtra

Experience

Where I've worked

Jr. Machine Learning Engineer

Anvex AI Technologies

Jul 2025 – Present · Vashi, India

  • Developed RAG-powered chatbot using Qdrant vector database, SambaNova Llama-4-Maverick-17B, and LangGraph, delivering context-aware answers across 22+ company policies
  • Built Doc2Dial automated voice platform using VAPI AI SDK and Llama-4-Maverick-17B for form-to-MCQ conversion, automated candidate calls, and evaluation report generation
  • Engineered audio processing pipeline with Sarvam AI STT and sentiment classifiers for Marathi conversations, reducing manual transcription and sentiment classification effort by 60%
  • Integrated key features for Anvex-Voice real-time voice AI system including Deepgram STT, GPT-4o-mini, RAG, and Cartesia Sonic TTS with streaming pipelines for partial response playback
  • Built a modular TTS middleware using the Factory Pattern to unify self-hosted endpoints and proprietary APIs like Sarvam, and deployed the containerized system on AWS EC2 for scalable voice generation.
  • Fine-tuned SLMs and TTS models using Unsloth on proprietary datasets for domain-specific optimization

Artificial Intelligence Intern

Eklavya.Me

Aug 2024 – Jun 2025 · Goa, India (Remote)

  • Engineered EkAIplatform using multi-agent orchestration with OpenAI GPT-4 and CrewAI, automating video creation and generating 100+ educational videos
  • Built Quiz Creator generating curriculum-aligned assessments in multiple formats (MCQ, true/false, short answer) using specialized task agents, serving 1,000+ students
  • Architected EkAIComic and CodeEkAI for automated comic strip and coding course generation using CrewAI and Stable Diffusion models
  • Researched and benchmarked 5+ text-to-image models (DALL-E, Stable Diffusion, Midjourney, Flux-Schnell) for consistent, high-quality generation
  • Containerized microservices with Docker, implemented CI/CD workflows using GitHub Actions, and authored technical documentation

Projects

Things I've built

Code Buddy (MCP Server)

Dec 2025 – Present

Developed open-source MCP server extending Claude Desktop with 29+ development tools for local file operations, git integration, Docker management, and HTTP requests. Architected async Python system with security layer and OpenAI API integration with real-time streaming responses.

Python 3.13MCP SDKOpenAI APIasyncioDocker
View on GitHub

SemanticCore

Sep 2025 – Oct 2025

Built custom transformer-based sentence embedding model from scratch using 4-layer encoder architecture with 8-head attention, 256-dim embeddings, and custom BPE tokenizer (10k vocab). Trained on 50k text samples using SimCSE contrastive learning, achieving 96.6% similarity on related sentences.

PyTorchTransformersCUDATokenizersSimCSE
View on GitHub

AI Voice Agent Platform

Sep – Nov 2025

Production-grade real-time TTS inference API with streaming capabilities for conversational AI agents. Sub-200ms latency with persistent WebSocket connections.

FastAPIPyTorchWebSocketDocker

CodeBridge

Oct – Sep 2025

AI-powered code context bridge for VS Code enabling seamless integration with AI platforms. Intelligent code chunking with 20+ language support.

FastAPIPythonWebSocketVS Code API

InstaNews

Mar – Apr 2025

Flask-based web application delivering personalized news reels using AI-powered summarization with Cohere API integration.

FlaskLangChainCohere APIPython

ContextualDoc

Jul – Aug 2024

Document querying system with multi-format support for context-aware Q&A using FAISS and Google Gemini.

PythonFAISSGoogle GeminiLangChain

Skills

Technologies I work with

Languages & Databases

PythonMySQLMongoDBPostgreSQL

AI & ML Frameworks

PyTorchTransformersLangChain/LangGraphCrewAIScikit-LearnUnslothvLLMOllama

Backend & APIs

FlaskDjangoFastAPIStreamlit

Vector Databases

PineconeQdrantFAISSChroma

DevOps & Tools

GitDockerKafkaSocket.IOAWS (EC2, S3, Lambda)Linux

Data Science

PandasNumPyMatplotlibSeabornPlotly

Education

Academic background

Bachelor of Engineering (B.E)

Artificial Intelligence And Data Science

New Horizon Institute Of Technology & Management

2021 - 2025CGPA: 9.20/10
  • Department Topper: Ranked 1st for semesters 1-5

Senior Secondary (XII)

Science - MSBSHSE

Bhavan's College

2020 - 2021Percentage: 85.71%

Contact

Get in touch

I'm always open to discussing new projects, creative ideas, or opportunities. Feel free to reach out.

© 2025 Abhishek Vishwakarma