We can organize an interview with Aldin or any of our 25,000 available candidates within 48 hours. How would you like to proceed?
Schedule Interview NowMy name is Abhilash V. and I have over 6 years of experience in the tech industry. I specialize in the following technologies: Deep Learning, Keras, Machine Learning, Natural Language Processing, TensorFlow, etc.. I hold a degree in , . Some of the notable projects I’ve worked on include: AI-Powered Video Semantic Search, AI-Powered Meeting Transcription & Speaker Diarization, OpenAI Experience, Product similarity search, Video Analytics, etc.. I am based in Trivandrum, India. I've successfully completed 6 projects while developing at Softaims.
My expertise lies in deeply understanding and optimizing solution performance. I have a proven ability to profile systems, analyze data access methods, and implement caching strategies that dramatically reduce latency and improve responsiveness under load. I turn slow systems into high-speed performers.
I focus on writing highly efficient, clean, and well-documented code that minimizes resource consumption without sacrificing functionality. This dedication to efficiency is how I contribute measurable value to Softaims’ clients by reducing infrastructure costs and improving user satisfaction.
I approach every project with a critical eye for potential bottlenecks, proactively designing systems that are efficient from the ground up. I am committed to delivering software that sets the standard for speed and reliability.
6 years
2 Years
2 Years
1 Year
Potentially possible
Google India
Developed an AI-powered video semantic search solution using advanced embedding models and FastAPI. The project focused on enabling precise frame-by-frame analysis of video content to extract meaningful scenes and segments based on user-defined prompts. By implementing state-of-the-art semantic search algorithms and efficient metadata indexing, the solution delivers fast and accurate results. Designed for applications like media indexing, content analysis, and personalized recommendations, this tool streamlines video retrieval processes and maximizes the value of video datasets.
A high-quality transcription and speaker diarization service tailored to your meeting recordings. With years of experience in natural language processing and AI solutions, I specialize in delivering accurate, time-efficient transcriptions that differentiate between speakers and capture every detail. Whether for business meetings, interviews, or conference calls, my service ensures clarity, reliability, and actionable insights. My work is powered by advanced AI tools, guaranteeing precision and quality, and I strive to exceed client expectations with a client-focused approach.
In this project, we aim to develop an advanced question-answering bot that leverages the power of OpenAI GPT-models and combines it with the Langchain framework and Pinecone DB to provide accurate answers using private data not indexed by default GPT models. Our solution incorporates the Langchain framework to enrich each API call to GPT-3.5/4 with relevant context, resulting in better outcomes without the need for model fine-tuning. To make the most of private data, we will employ Pinecone DB , and its approximate nearest neighbor (ANN) search capabilities for efficient text indexing and retrieval of similar contexts. By combining these technologies, the question-answering bot will be able to provide more precise answers without fine-tuning, taking advantage of the vector similarity and vector database to enhance the context length of each query. This cutting-edge solution is perfect for businesses looking to harness the power of AI for efficient and accurate information retrieval using their unique, private data.
We can search through a database of products and find visually similar products. This feature can be used to do product deduplication on an eCommerce platform and to find the same product across multiple sites for tasks like price comparison. This can be extended to add text features and metadata features like price and type etc to make it more robust in matching the product.
Video analytics done for retail stores with face detection, face recognition, people counting, action recognition and heat map generation with activity level of the floor. Multiple video feeds are consumed and divided into image frames and send to processing modules which would process these frames and detect persons, faces and actions. This is a modular architecture as we can add more processing modules to get more insights from the video data.
in Deep Learning
in
2011-01-01-2015-01-01