Sushant kumar

Software Engineer

Bengaluru, Karnataka, India1 yr 3 mos experience
AI EnabledAI ML Practitioner

Key Highlights

  • Expert in Generative AI and RAG pipelines
  • Designed a digitization platform for ancient manuscripts
  • Delivered 6 projects simultaneously with high efficiency
Stackforce AI infers this person is a specialist in AI/ML solutions for document processing and multimedia applications.

Contact

Skills

Core Skills

Llm IntegrationRag PipelinesComputer VisionAi/mlNlp

Other Skills

Neural NetworksLangChain (RAG)OCRVertex AIREST APIsChromaDBTensorFlowArtificial Intelligence (AI)Machine LearningDeep LearningChatbotsKerasSentence TransformersHTML/CSS/JavascriptFlask

About

I am a Software Engineer at Bluerose Technologies with a B.Tech in Computer Science. I have 1.7 years of experience specializing in Generative AI and RAG pipelines, delivering 4–6x industry-average output.Expert in LLM integration, local AI deployment, and multi-project technical leadership. Proven track record: 6 projects simultaneously, 40–60 person-days of work delivered in ~10 working days. I build AI systems that don’t just work on modern data — but on centuries-old manuscripts, degraded documents, and low-resource scripts. My focus is at the intersection of custom OCR, LLM-based post-correction, and scholar-in-the-loop workflows. Recently, I designed and implemented a complete digitisation platform for Hale Kannada (Ancient Kannada) manuscripts — a domain where standard OCR fails. The system uses custom-trained OCR (90%+ target accuracy), LLM correction with Jain/Sanskrit vocabulary constraints, script conversion (Kannada → Devanagari → Telugu), and a side-by-side reviewer interface. It turns scanned PDFs into searchable, reusable digital text. Outside of manuscripts, I’ve built production RAG pipelines, photo restoration systems, cinematic lyric-video generators, and OCR infrastructure for 20,000+ books — all with a strong bias toward cost-efficient, deployable, real-world AI. I design multilingual knowledge retrieval platforms using Vertex AI, LangChain, and vector databases like Chroma.I build robust web interfaces and REST APIs using Python. Expert in LLM Integration, RAG Pipelines, Prompt Engineering, LangChain, OCR, and Computer Vision. Core Tech Stack: Languages: Python, OOPS. AI/ML: RAG, LLMs, Vertex AI, TensorFlow, PyTorch, and NLP, computer vision, OCR Tools: Pytorch, Docker, TensorFlow,REST APIs, Sentence Transformers, and Agile Scrum. I am a Salesforce Certified AI Associate, Meta Certified Full-Stack Engineer, and Google Certified Project Management Professional. I am driven by the goal of building intelligent systems that solve real-world problems

Experience

1 yr 3 mos
Total Experience
1 yr 3 mos
Average Tenure
1 yr 3 mos
Current Experience

Bluerose technologies

2 roles

Software Engineer

Mar 2025Present · 1 yr 3 mos · Bengaluru · On-site

  • Built Hale Kannada OCR + LLM correction platform for ancient manuscript digitisation: custom Kraken OCR model → LLM refinement (temperature 0, Jain/Sanskrit lexicon) → script conversion (Kannada, Devanagari, Telugu) → scholar review interface → multi-format export (JSON, TXT, PDF).
  • Engineered 3-stage AI photo restoration pipeline (Phi-4 → Flux Kontext → Real-ESRGAN) on Supabase Edge Functions → 42% cost reduction, sub-1-min per image.
  • Developed 7-stage multilingual book OCR pipeline (Chandra OCR + Groq LLM + 4-language translation) → 168 semantic chunks, REST API integrated live.
  • Built automated cinematic lyric-video pipeline (Whisper timestamps → Llama 3 prompts → Flux-Pro 1920×1080) → delivered 500 production videos.
  • Proposed architecture reducing 20,000-book OCR library cost from $0.052 → $0.030 per image (42% saving).
  • Designed enterprise RAG system with ChromaDB + Vertex AI → embedding, retrieval, reranking, citation-aware generation for multilingual documents.
  • Built LLM personalisation engine with FSRS + ELO-based adaptive learning.
  • Integrated high-throughput WhatsApp API with async queue → multilingual NLP intent extraction.
Neural NetworksLangChain (RAG)OCRLLM IntegrationRAG PipelinesVertex AI+1

Software Trainee

Oct 2024Mar 2025 · 5 mos · Bengaluru · On-site

  • Built web-scraping + OCR pipeline on 50,000+ records → reduced manual documentation by 30%.
TensorFlowArtificial Intelligence (AI)AI/ML

Education

Darbhanga College of Engineering

B. Tech — CSE

Sep 2020Aug 2024

Ambika bhawani public school

Intermediate — PCM

Jul 2018Jun 2020

Jawahar Navodaya Vidyalaya - JNV

Secondary education

Aug 2013May 2018

Stackforce found 100+ more professionals with Llm Integration & Rag Pipelines

Explore similar profiles based on matching skills and experience

Sushant kumar - Software Engineer | Stackforce