H

Harshit Preetam R

About

Detail

AI Engineer specializing in Large Language Models, Computer Vision, and Edge AI. I focus on designing intelligent systems that learn, adapt, and perform reliabl
Hyderabad, Telangana, India
Contact Harshit regarding: 
Flexible work
Starting at USD12/hour

Timeline

work
Job
school
Education
folder
Project
flag
Award

Résumé

Jobs verified_user 0% verified
  • S
    AI Engineer
    SSEV SoftSols Pvt Ltd
    Jan 2025 - Current (1 year 9 months)
    – Built and deployed production-focused AI systems in Python with scalable inference pipelines and backend orchestration.
    – Optimized transformer-based and deep learning models to improve accuracy, reduce inference latency, and ensure efficient
    performance under production-level workloads.
    – Designed LLM-powered RAG workflows using embeddings, retrieval strategies, and structured prompting for context-aware AI
    applications.
    – Built modular FastAPI-based AI microservices with structured logging, validation, and asynchronous task handling using Redis and
    Celery to support scalable model serving and seamless backend integration.
    – Built multi-step AI workflows with retrieval pipelines, context management, and
Education verified_user 0% verified
  • Mahindra University
    B.Tech in Artificial Intelligence
    Mahindra University
    Aug 2021 - Aug 2025 (4 years 1 month)
Projects (professional or personal) verified_user 0% verified
  • A
    AlphaOPT Enhancement Pipeline public Remote experience
    Feb 2026 - Current (8 months)
    – Enhanced an autonomous optimization code-generation pipeline using local LLMs to solve mathematical programming tasks from
    the NL4OPT benchmark through executable Gurobi solver generation.
    – Implemented persistent failure-memory logging and confidence-based retrieval while resolving structured output, retrieval, encoding,
    and malformed JSON failures.
    – Improved Pass@1 from 54.7% to 59.4%, increased execution success from 89.1% to 92.2%, and reduced evaluation runtime from 62
  • A
    AI OCR Answer Evaluation System
    Jul 2025 - Aug 2025 (2 months)
    - Developed an AI-powered exam grading platform that processes handwritten answer booklets using TrOCR OCR and automatically evaluates responses.
    - Implemented multi-step AI grading workflows combining semantic similarity, numerical tolerance, and symbolic algebra evaluation.
    - Designed scalable FastAPI services and APIs with asynchronous grading pipelines powered by Redis and Celery. Additionally, to enhance user experience, I recognized the need for a dashboard that allows users to easily access and test the grading system, which ties into my understanding of full-stack development principles, even though my primary focus was on back-end services. This approach ensured that the platform was user-friendly and accessible, ultima
  • N
    NFC Payment System
    Mar 2025 - Jun 2025 (4 months)
    Built a cloud-integrated NFC payment platform with merchant and customer dashboards, real-time wallet sync, and sub-1s latency transactions. Integrated Supabase for scalable storage and an LLM-powered assistant for transaction insights and query handling.
  • M
    Multimodal AI Agent
    Nov 2024 - Feb 2025 (4 months)
    Developed a multimodal agent integrating voice (Whisper) and gesture (Mediapipe) inputs for edge automation on Jetson hardware, leveraging cloud inference for high-accuracy, real-time control. Achieved fully offline inference with 90% command accuracy.
  • A
    Autonomous Ball Detection & Dropping Robot
    Jan 2024 - Aug 2024 (8 months)
    Designed a vision-guided robot for competitive ball sorting, integrating YOLOv8 for real-time detection with sub-50ms inference and optimized control for high-speed, precision drops in dynamic environments. Achieved 80% drop accuracy in dynamic competition conditions.
  • T
    Twitter Sentiment Analysis
    Nov 2023 - Dec 2023 (2 months)
    Engineered a live sentiment monitoring pipeline for Twitter data, utilizing natural language processing (NLP) and transformer embeddings to effectively track public opinion. Processed over 1,000 tweets per day and generated visual analytics through word clouds and trend plots, enhancing the interpretability of sentiment trends. Additionally, developed a comprehensive dashboard for visualization, incorporating back-end development techniques to ensure seamless data integration and user experience. Although HTML was not a primary focus, its application in the dashboard design contributed to a more interactive and visually appealing interface, ultimately optimizing the presentation of analytical insights. Leveraged machine learning frameworks
Awards verified_user 0% verified
  • A
    Awarded 1,00,000 merit scholarship in the first year of B.Tech.
    Aug 2021 - Current (5 years 2 months)