Akshay Joshi

Senior AI Research Scientist at BlinkIn

Previously: AMD, Google, ETH Zurich, DFKI, and Saarland University.

Know More Contact Me

Technical Skills

CV-backed core stack for AI research and ML engineering roles in 2026

Model Engineering

Python

Primary language for model training, evaluation pipelines, and production AI systems.
Core Language
Model Engineering

PyTorch

Core deep learning framework for large-scale LLM and multimodal model development.
Foundation Framework
Model Engineering

Transformers (Hugging Face)

Transformers, Tokenizers, and Accelerate stack for foundation model development.
Foundation Models
Research

LLM/VLM Fine-Tuning (PEFT, LoRA, SFT)

Parameter-efficient adaptation of language and vision-language models for domain tasks.
High Demand 2026
Research

RLHF & DPO

Alignment pipelines to reduce hallucination and improve instruction-following quality.
Alignment Stack
Research

Multimodal AI (Vision-Language + Video)

Vision-language modeling, cross-modal reasoning, and video understanding systems.
VLM + Video
Production

RAG & Semantic Search

Retrieval-augmented generation and semantic retrieval for knowledge-intensive AI products.
Enterprise AI
Production

Vector Retrieval (Qdrant, Milvus, FAISS)

Scalable vector indexing and retrieval layers for search and recommendation workloads.
Retrieval Infra
Production

Agentic AI & MCP (LangGraph/LangChain)

Tool-using AI workflows with LangGraph, LangChain, and Model Context Protocol.
Agent Workflows
Model Engineering

Inference Optimization (vLLM, TensorRT, DeepSpeed)

Serving and optimization stack for throughput, memory efficiency, and response latency.
Low-Latency Serving
Production

FastAPI + Docker + Kubernetes

Model-serving APIs and container orchestration for production AI deployment.
Deployment Stack
Production

Cloud/Data/Experiment Stack

AWS/GCP/Azure, SQL/Elasticsearch/Redis, MLflow/W&B, and Git/CI-CD workflows.
Scale & Ops

Publications

Seminars, Conferences & Publications that I have delivered/currently working on

Art Style Classification with Self-Trained Ensemble of AutoEncoding Transformations

Pattern Recognition & Computer Vision
Investigation on the use of deep semi-supervised neural models to extract dense features in complex & ambiguous images spanning across 27 unique artistic styles. Self-supervision enforced to resolve class imbalance of WikiArt dataset
PDF Code Dataset Source Document
Learning like humans with Deep Symbolic Networks

Seminar
Amalgamating the reasoning power of Symbolic AI with the learning ability of Deep Neural Networks as part of Hybrid Machine Learning Approaches and Applications Seminar
PDF Slides Source Document
Research Posters on Advanced Cryptography & High Performance Quantum Computing

Computer Society of India
At Computer Society of India 2017 conference, presented a research poster on Advanced Cryptographic Standards & Security. Further, delivered seminars on Quantum Computing & Phased Array Antennas for 5G Mobile Applications
Source Document

Projects

Featured projects spanning across different research areas in Vision, NLP & Data Mining

*
Message Queuing Telemetry Transport Client

Message Queuing Telemetry Transport Client

MQTT Client application to simulate lightweight, publish-subscribe network protocol for IoT/Edge devices.

File Transfer Simulation with Integrity Check

File Transfer Simulation with Integrity Check

Socket Programming using Python, TCP & Hashing

Graph Convolution Network Classifier with Sparse Adjacency Matrix Prior

Graph Convolution Network Classifier with Sparse Adjacency Matrix Prior

Benchmarking FCN, CNN & GCN Classfier with a Custom Prior

COVID-19 Tweets Sentiment & Exploratory Data Analysis

COVID-19 Tweets Sentiment & Exploratory Data Analysis

Deep Self-supervised Transfer Learning based Sentiment Analysis of COVID19 Tweets

Robust Spoken Language Recognition

Robust Spoken Language Recognition

Uncover hidden patterns in the phoneme vector space by performing Pairwise Cosine Similarity, Dimensionality Reduction & Spacial Clustering

Multi-stage Information Retrieval & Ranking System

Multi-stage Information Retrieval & Ranking System

Search Engine implemented using Language Models, BM25 (Okapi, Plus) & TF-IDF

Open-domain Question Answering with End-to-End Memory Networks

Open-domain Question Answering with End-to-End Memory Networks

Deep neural implementation of a QA system using Knowledge Graphs & Wikipedia Corpus

Event-driven Pedestrian & Vehicle Detection

Event-driven Pedestrian & Vehicle Detection

Implementation of a real-time ROI-based Vehicle Detection and Counting system using OpenCV & YOLO

Context-aware Intelligent Assistant with Deep Q Learning

Context-aware Intelligent Assistant with Deep Q Learning

Devising a reinforcement learning based architecture for robust speech assistants

HONOURS & ACHIEVEMENTS

ACCOMPLISHMENTS THAT BOLSTERED MY ACADEMIC & LEADERSHIP FINESSE

Professional Recommendations
Advanced Micro Devices Feb 2018 – Sep 2019

 

  • Recommended by the Director of Technology & Software Engineering and various Senior Engineers at AMD.
  • Demonstrated solid technical expertise and agile adaptability skills.

Bachelor of Engineering Class Rank
Visvesvaraya Technological University Aug 2013 – Jun 2017

 

  • Ranked in the Top 5% of the graduating class in Computer Science & Engineering department.
  • Total number of students in the graduating cohort of 2017 were 106.

Academic Recommendations
Visvesvaraya Technological University Aug 2013 – Jun 2017

 

  • Recommended by Prof. Shiva Kumar Dalali, Prof. Manjunatha P.B and Prof. Raghavendra T.S in the Computer Science department.
  • Illustrated exceptional research/course work in Discrete Mathematics, Computer Graphics & Visualization.

Vice-Captain
Pramana Pre-University College Dec 2012 – Dec 2012

 

  • Vice-Captain of state-level Throwball team, representing Raichur district, India.
  • The Pre-University tournament was held at Mysore, Karnataka, India.

Contact

Connect with me

Email Call Book an appointment

Map loads on demand for a faster mobile experience.