Rachit R Jindal

Experience

• 

  Nirmitee.io

  Dec 2021 - Feb 2022

  Machine Learning Enginner

  1. Worked on Topic Modelling using BERTopic for clustering the content/abstract with their relevant or similar topic based on the trained data.2. Developed QnA machine learning model using 'question-answering' pipeline of Transformers.3. Developed API of Content-Based Product Recommendation Engine using Flask, Tensorflow, MongoDB, and Sklearn

• 

  Disecto

  Mar 2022 - Jun 2022

  Software Developer

  Worked on PII extraction pipeline and multi-lingual OCR.

• 

  Fasthr AI

  Dec 2022 - Dec 2023

  • Build and deploy ML microservice with RabbitMQ for async communication improving the uptime for ML service by about 70% and improving the product's scalability.• Built a custom ML computer vision pipeline using Yolov8, SAM, and FaceClassifier for real-time proctoring using AWS ECS, S3, and RabbitMQ• Integrated LLM orchestration into production using function calling in GPT-4, for channeling content to specific LLM like Claude-2, Llama, and GPT-4 based on actions optimizing cost and context window• Specialized in working with vector databases for production applications.• Implemented a weight-sum algorithm for re-ranking content retrieved from vector databases based on user search criteria.• Successfully executed multiple POC to validate and demonstrate the feasibility of different solutions.• Contributed to CI/CD pipelines, ensuring efficient and automated deployment processes.• Developed High-Level Designs (HLD) for entire HR tech moat applications. Show less - Developed SaaS products for natural language processing using generative AI, focusing on prompt engineering for fine-tuning GPT-3.5-turbo.- Utilized RabbitMQ as a message broker for efficient backend communication and AI container synchronization.- Deployed a RESTful API into production on Azure and AWS to facilitate seamless integration within SaaS products.- Implemented cutting-edge research papers in Generative AI, including Generative AI Agents, MemPrompt, Guide Rail, enhancing performance and functionality.- Applied the reflection technique to control hallucinations of large language models (LLM) in a conversational AI platform.- Successfully deployed the recruitment automation SaaS tool Fasthr.ai, leveraging GPT-4 to streamline and enhance the recruitment process. Show less

  • AI Platform Solutions Engineer

    Jul 2023 - Dec 2023

  • Generative AI Engineer

    Dec 2022 - Jul 2023
• 

  Brane Group

  Jan 2024 - Sept 2024

  AI/ML Engineer | Associate Solution Leader

  • POC on retail price optimization using DQN Reinforcement learning to maximize margins and revenue with a custom environment.• Developed ML-backed microservice architecture for predictive analytics tool, with high uptime and scalability using pub-sub.• Build Retrieval-augmented generation (RAG) using hybrid search for efficient text retrieval.• Performed benchmarking on custom data of multiple LLM using Meteor, ROUGE-L, and BLUE scores.• POC on dynamic NER using GLiNER to extract entities from multiple docs on user-provided labels.• RoBERTa, and DeBERTa for text sentiment analysis. Show less

• 

  8bit.ai

  Sept 2024 - now

  Sr. AI Engineer

  - Researching and developing interference engine for deployment of LLM models at scale with high GPU utilisation and efficient performance.- Worked on implementing and testing workflows for deploying Server-less LLM with autoscaling for online inference- Optimised scheduling system with priority based recursive fallbacks on batch requests - Implemented statical logic for optimising LLM deployments by dynamically determining model footprint, and compute requirements for KV cache, TTFT, and TPOT calculations for optimum tensor parallel and pipeline parallel parameter- Benchmarked inference engines like vLLM, SgLang, TGI. - Contributed in development of RAG architecture involving custom retrieval algorithm combining Query Expansion + Hybrid search + ReRanker + custom sorting algorithm with optimised vectordb operations for reducing latency- Experimented on BareMetal infra for deploying Kuberay operator for automating GPU discovery to RayCluster via gpu-feature-discovery Show less