Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

I am Bharath Sai Reddy Avuthu, an AI Engineer focused on building and deploying production-grade LLM and computer vision systems. I specialize in quantized inference pipelines, multi-agent orchestration (LangGraph), and deterministic control layers (FSM) to manage edge cases and minimize model failure modes. I place a strong emphasis on reliability, latency optimization, and domain-aware decision systems for real-world deployments. I enjoy turning complex problems into robust, scalable solutions and thrive in environments that value reliability and thoughtful engineering for real-world impact.…I am Bharath Sai Reddy Avuthu, an AI Engineer focused on building and deploying production-grade LLM and computer vision systems. I specialize in quantized inference pipelines, multi-agent orchestration (LangGraph), and deterministic control layers (FSM) to manage edge cases and minimize model failure modes. I place a strong emphasis on reliability, latency optimization, and domain-aware decision systems for real-world deployments. I enjoy turning complex problems into robust, scalable solutions and thrive in environments that value reliability and thoughtful engineering for real-world impact.

Bharath Sai Reddy Avuthu





I am Bharath Sai Reddy Avuthu, an AI Engineer focused on building and deploying production-grade LLM and computer vision systems. I specialize in quantized inference pipelines, multi-agent orchestration (LangGraph), and deterministic control layers (FSM) to manage edge cases and minimize model failure modes. I place a strong emphasis on reliability, latency optimization, and domain-aware decision systems for real-world deployments. I enjoy turning complex problems into robust, scalable solutions and thrive in environments that value reliability and thoughtful engineering for real-world impact.…I am Bharath Sai Reddy Avuthu, an AI Engineer focused on building and deploying production-grade LLM and computer vision systems. I specialize in quantized inference pipelines, multi-agent orchestration (LangGraph), and deterministic control layers (FSM) to manage edge cases and minimize model failure modes. I place a strong emphasis on reliability, latency optimization, and domain-aware decision systems for real-world deployments. I enjoy turning complex problems into robust, scalable solutions and thrive in environments that value reliability and thoughtful engineering for real-world impact.

Available to hire

I am Bharath Sai Reddy Avuthu, an AI Engineer focused on building and deploying production-grade LLM and computer vision systems. I specialize in quantized inference pipelines, multi-agent orchestration (LangGraph), and deterministic control layers (FSM) to manage edge cases and minimize model failure modes. I place a strong emphasis on reliability, latency optimization, and domain-aware decision systems for real-world deployments.

I enjoy turning complex problems into robust, scalable solutions and thrive in environments that value reliability and thoughtful engineering for real-world impact.

Skills

Experience Level

Expert

Expert

Expert

Expert

Language

English

Advanced

Work Experience

Data Scientist at Endovision

June 1, 2024 - Present

Led edge quantization to 8-bit PTQ with layer-wise sensitivity analysis, maintaining accuracy within 1% of FP32. Implemented history-aware confidence aggregation (k-frame window) reducing production false positives by 10–15%. Matched EfficientNet-B7 performance using EfficientNet-B0, and improved multi-label F1 from 70 to 80+. Expanded object detection with two new classes without metric degradation. Implemented dynamic runtime switching between FP16 and quantized models to sustain <50 ms latency under hardware constraints. Refactored ensemble decision logic into matrix-vector operations, reducing decision latency by 7 ms. Engineered a heuristic-based separation layer to add a new class without full retraining.

Data Science Intern at Endovision

January 1, 2024 - June 1, 2024

Developed a modular training pipeline separating model architecture from configuration, reducing experimentation cycle time and enabling structured hyperparameter tracking. Implemented version-controlled experiment tracking to monitor and reproduce model runs across iterations.