Suyash Bhatkar

Suyash Bhatkar

About

Detail

AI Developer | Agentic AI | NLP | Machine Learning/Deep Learning | SQL | AI/MLOps | CI/CD | AWS
Navi Mumbai, Maharashtra, India
Contact Suyash regarding: 
Flexible work
id_card
Internships

Timeline

work
Job
school
Education
folder
Project

Résumé

Jobs verified_user 0% verified
  • AP Guru
    AI Developer
    AP Guru
    Mar 2026 - Current (6 months)
    Architecting an AI-native tutoring platform on FastAPI with a multimodal RAG pipeline (Pinecone + OpenAI text-embedding 3-small + BM25 hybrid retrieval) and an LLM function-calling layer over Gemini. Grounded responses in real student data via async tool execution, SSE streaming, and Langfuse-traced observability. Designing reusable ingestion/indexing, semantic-retrieval, and adaptive quiz-generation frameworks (strategy-pattern question selection, per-test AI analysis, auto-tagged sessions) enabling continuous data expansion, personalized study plans, and rapid rollout of new AI-driven features across the SAT learning ecosystem.
  • N
    AI Product Intern
    Navneet Toptech
    Nov 2025 - Jan 2026 (3 months)
    Engineered a Curriculum Comparison System (Ensemble Agreement Model) that reduced comparison time from 30 days to 10 minutes for cross-board curricula. Developed an OCR (EasyOCR) pipeline for automatic classification of exam papers (grade/subject), enabling digital indexing of historical data with custom trained YOLO models for question extraction. Built a Lesson Plan Generating Agent, where user uploads a chapter and a lesson plan of how to teach students as a guidance for teachers is created for NTT's LMS platform.
  • B
    AI Engineering Intern
    Breakout Escape Rooms
    Sep 2025 - Oct 2025 (2 months)
    Led a team of 3 interns to design and develop an end-to-end AI voice booking assistant that handled customer calls using Elevenlabs, Bookeo, PayU and Supabase as database. Built and integrated FastAPI routers to connect backend-frontend, ensuring communication across modules. Integrated LLM-based summarization workflows into operations, increasing support efficiency by ~40%. Implemented scalable data handling and a system used by 100+ people on a monthly basis.
Education verified_user 0% verified
  • University of Mumbai
    Bachelor of Science, Physics
    University of Mumbai
    Jan 2024
    Ramnarain Ruia Autonomus College, Matunga
Projects (professional or personal) verified_user 0% verified
  • p
    Vaani
    peronal
    Apr 2026 - Current (5 months)
    A full fledged AI powered, open source, self deployable, BYOK model for AI Powered Finance Tracking and expense analysis.
  • AP Guru
    AI Tutor Chat
    AP Guru
    Mar 2026 - Current (6 months)
    Function Calls to give AI the power of data, and implementation of per user memory using a state json. Utilizing a routing model, to assign related tasks like basic query solving or high level question solving and cost optimization while keeping quality of output high.
  • F
    FINE TUNED TRANSFORMER BASED HINGLISH - ENGLISH TRANSLATION
    Sep 2025
    Built a two-stage multilingual pipeline using Indictrans (Hinglish→Hindi) + MarianMT (Hindi→English) model. Preprocessed & normalized and fine-tuned the model on the 100k+ Hindi English Paralell Corpora. Achieved BLEU score of ~85% and a BERT Score of ~90% Deployed on HuggingFace Spaces
  • F
    FINE-TUNING OF LLM MODEL (GOOGLE'S GEMMA 2B) USING LORA/QLORA
    Sep 2025
    Fine-tuned the LLM Gemma-2B for legal text summarization using LoRA + 4-bit quantization (QLoRA) to enable training on limited GPU memory. Built a complete training & evaluation pipeline with Hugging Face transformers, datasets, and peft. Optimized performance while achieving ~4x GPU memory savings compared to full fine-tuning.
  • A
    ADVANCED RAG PROTOTYPE WITH HYBRID SEARCH AND RERANKING
    Sep 2025
    Built and deployed a hybrid RAG pipeline integrating LangChain, ChromaDB, and BM25. Reranking mechanism and integrated Groq API via LangChain, improving response relevance by ~30% Deployed on Hugging Face Spaces with a scalable architecture roadmap for AWS/GCP, demonstrating end-to-end LLM application development skills.