Hrushik Reddy

hrushik reddy

ai engineer.

India-based ai engineer, currently at techolution. for the past 2+ years i've been fine-tuning open-source llms, building real-time speech pipelines, and designing the infra that keeps them alive in production. i care about systems that actually ship - demos are easy, uptime is the hard part.

about

tldr; i'm the person who takes a model from “works in a notebook” to “serves 20k+ requests a day without falling over”. at techolution i fine-tune and serve models like llama-3, deepseek and qwen-vl (~35% accuracy gains, ~40% lower latency), and build speech pipelines that go from mic to voice in under 2 seconds.

before this i shipped vision-llm content moderation at multeway and an adaptive-rag llama-2 tutor at eklavya. i mostly live in python, fastapi, pytorch, and whatever serving stack keeps p99 latency down. i also build small things after hours - terminal apps, cli helpers, on-device asr for telugu.

when i'm not shipping models i'm writing about the build, grinding leetcode, or taking something apart to see how it works. open to hard ai problems, collaborations, and conversations - i reply unreasonably fast. grab my resume or say hi at phanihrushik.10@gmail.com.

projects

exp.

stack

genai & llm systems
multi-agent (adk, langgraph), fine-tuning & lora, quantization, rag pipelines, serving optimization
ml & deployment
pytorch, computer vision, chromadb / faiss / pgvector, prometheus + grafana, model evaluation
backend & apis
python, fastapi, rest apis, spring boot, sse streaming
cloud & distributed
gcp, aws (s3, sqs), docker & kubernetes (helm), kafka / rabbitmq / celery, serverless gpus
frontend
react, next.js, angular, typescript, html / css
speech & edge ai
whisper (cloud to edge), kokoro tts, web audio api, whisper.cpp on arm64, android ndk / jni

credentials

osmania university8.41/10 · 2025

be, electronics & communication engineering

certifications

achievements

writing