Headline

AI Developer Intern @Forage AI || AI Engineer · LLMs · RAG · Agentic AI · LangChain · Full-Stack AI Systems | IIITDM Kurnool AI&DS’27

About

I build AI systems that go beyond demos, production pipelines, real deployments, measurable results. At IIITDM Kurnool (Branch Rank #2, CGPA 8.9), I specialize in Generative AI, LLM-powered applications, and distributed systems. Over the past year I've shipped 6+ end-to-end AI projects spanning RAG pipelines, multimodal video reasoning, computer vision, and full-stack agentic platforms, each one solving a real problem for real users. A few things I've built and what they achieved: → Extended the MR.Video MapReduce framework for long-form video understanding — 49% fewer API calls, 48% token reduction, 24% latency drop, +10% accuracy on a dual-node CPU cluster → Croporia — a 3-service AI agricultural platform (React + FastAPI + LangChain + LLaMA-3.3-70B) serving 43+ crop profiles, 7 live APIs, and a RAG chatbot for Indian farmers → SevaSetu — a multilingual AI government assistant (AWS Bedrock Claude 3.5 Sonnet, 9 Indian languages, EasyOCR) live at sevasetu-demo.vercel.app → AbleEat — ResNet + Vision + LLM shelf-scanning system, 98% classification accuracy, 30% unsafe-item recall improvement Technically, I work across the full AI stack: LangChain, CrewAI, AutoGen, RAG, MCP, TensorFlow, FastAPI, React, AWS (Bedrock, SageMaker, EC2, S3), Docker, and PostgreSQL. If you're working on hard problems at the intersection of LLMs, distributed systems, or real-world AI deployment — let's talk.

Location

Vijayawada

AI stack

cursor

opencode

Work experience

AI Developer Intern at Forage AI, Jun 2026 - Present · 3 mos, New York, United States

Machine Learning Internship at FlyRank AI, Jun 2026 - Present · 3 mos, Vijayawada

Placement Coordinator at Indian Institute of Information Technology Design & Manufacturing, Kurnool, Aug 2025 - Present · 1 yr 1 mo

AI Intern · MelodAI - AI Music Generation System at Infosys Springboard, Nov 2025 - Jan 2026 · 3 mos, Remote, Engineered a 3-stage LLM-guided music generation pipeline combining Meta MusicGen with GPT-4 prompt refinement, producing 50+ audio samples across 8 mood and style categories with measurable stylistic consistency. Designed a 20+ parameter Flask inference API achieving sub-500ms average response time; built a React + Tailwind frontend with 10+ interactive generation controls. Implemented CI/CD-backed deployment across 4 Agile sprints; structured prompt templates that improved mood-to-audio alignment consistency by ~22% over unstructured baselines. Isolated nondeterministic LLM modules behind deterministic service boundaries, preventing prompt variance from impacting API contract stability.

Artificial Intelligence Intern at Samsung Innovation Campus, Dec 2024 - Mar 2025 · 4 mos, Built an LSTM-based speaker recognition model trained on 300+ audio samples; designed an MFCC preprocessing pipeline (13 coefficients, 12+ hyperparameter configurations) achieving 84% test accuracy, 18% above the untreated baseline. Noise-reduction preprocessing reduced feature variance by 25%, improving model robustness across varied recording conditions. Completed 200+ hours of structured ML/DL coursework spanning supervised learning, neural networks, and Generative AI fundamentals; deployed a real-time Streamlit inference interface with sub-200ms prediction latency.

Education

Bachelor of Technology - BTech, Indian Institute of Information Technology Design & Manufacturing, Kurnool

AAKASH INSTITUTE

Public links

github: https://github.com/Lokesh-916