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

Hi, I’m Akhileswar Pathinavalasa, an AI/ML engineer based in Texas, USA, with 5+ years of experience building and scaling ML systems and LLM-powered applications. I specialize in generative AI, retrieval-augmented generation, and distributed ML, and I’ve delivered production-grade solutions that improved accuracy and reduced latency in large-scale environments. I love partnering with product teams to ship impactful features using Python, PyTorch, and cloud platforms like AWS and Azure. I’m passionate about turning complex data problems into reliable, scalable systems and continuously optimizing model performance in real-world settings.…Hi, I’m Akhileswar Pathinavalasa, an AI/ML engineer based in Texas, USA, with 5+ years of experience building and scaling ML systems and LLM-powered applications. I specialize in generative AI, retrieval-augmented generation, and distributed ML, and I’ve delivered production-grade solutions that improved accuracy and reduced latency in large-scale environments. I love partnering with product teams to ship impactful features using Python, PyTorch, and cloud platforms like AWS and Azure. I’m passionate about turning complex data problems into reliable, scalable systems and continuously optimizing model performance in real-world settings.

Akhileswar Pathinavalasa

AI Engineer, Data Scientist, Developer, +2





Hi, I’m Akhileswar Pathinavalasa, an AI/ML engineer based in Texas, USA, with 5+ years of experience building and scaling ML systems and LLM-powered applications. I specialize in generative AI, retrieval-augmented generation, and distributed ML, and I’ve delivered production-grade solutions that improved accuracy and reduced latency in large-scale environments. I love partnering with product teams to ship impactful features using Python, PyTorch, and cloud platforms like AWS and Azure. I’m passionate about turning complex data problems into reliable, scalable systems and continuously optimizing model performance in real-world settings.…Hi, I’m Akhileswar Pathinavalasa, an AI/ML engineer based in Texas, USA, with 5+ years of experience building and scaling ML systems and LLM-powered applications. I specialize in generative AI, retrieval-augmented generation, and distributed ML, and I’ve delivered production-grade solutions that improved accuracy and reduced latency in large-scale environments. I love partnering with product teams to ship impactful features using Python, PyTorch, and cloud platforms like AWS and Azure. I’m passionate about turning complex data problems into reliable, scalable systems and continuously optimizing model performance in real-world settings.

Available to hire

Hi, I’m Akhileswar Pathinavalasa, an AI/ML engineer based in Texas, USA, with 5+ years of experience building and scaling ML systems and LLM-powered applications. I specialize in generative AI, retrieval-augmented generation, and distributed ML, and I’ve delivered production-grade solutions that improved accuracy and reduced latency in large-scale environments.

I love partnering with product teams to ship impactful features using Python, PyTorch, and cloud platforms like AWS and Azure. I’m passionate about turning complex data problems into reliable, scalable systems and continuously optimizing model performance in real-world settings.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Intermediate

Work Experience

AI/ML Engineer at Cursor

July 1, 2025 - Present

Architected a retrieval-augmented generation (RAG) pipeline to improve code suggestion accuracy by 18% across 1.2M daily completions for 40K+ active users. Reduced model inference latency by 40% (320ms to 190ms) via embedding-layer optimization and batched request handling, enabling a 25% increase in concurrent sessions. Fine-tuned a custom transformer on 8B+ tokens of multilingual code corpora, raising internal pass@1 from 62% to 74%. Built an A/B testing framework shortening experiment cycles from 3 weeks to 5 days, accelerating iteration on completion quality. Co-shipped a context-aware autocomplete feature, contributing to a 12% rise in weekly user retention in the first quarter.

ML Engineer Intern at Synthesia

December 1, 2024 - May 1, 2025

Developed a lip-sync alignment module reducing audio-visual drift by 35% and improving human evaluation scores from 3.6 to 4.2/5.0. Processed 14TB of video-audio training data via Apache Airflow, cutting data prep time from 9 days to under 3. Integrated OpenCV-based face detection to filter 22% of low-quality training frames, yielding sharper generated video. Achieved 85% unit and integration test coverage, reducing production incidents tied to model updates by 40%.

Data Scientist at JPMorganChase & Co.

November 1, 2021 - July 1, 2023

Built an XGBoost credit-default model on 4M+ loan records (AUC 0.91), saving an estimated $3.8M annually in early-stage default losses. Automated a 60+ analyst-hour monthly portfolio risk report into a 4-hour Python/Airflow pipeline — a 93% reduction in manual effort. Designed an NLP classification system to auto-tag 150K+ support tickets per quarter with 89% accuracy, freeing 15 FTEs from manual triage. Migrated 6 legacy SAS models to Python/scikit-learn with full metric parity, 3 weeks ahead of schedule. Presented model-risk findings quarterly to a 12-member senior review board, securing production approval for 3 new trading system models.

Reporting Analyst at Deutsche Bank

September 1, 2019 - October 1, 2021

Built 25+ SQL-driven Tableau dashboards tracking P&L, trade volumes, and counterparty exposure—eliminating 40 hours/week of manual Excel reporting. Investigated a $1.7M reconciliation discrepancy across fixed-income records by cross-referencing 3 internal data sources, triggering a process overhaul. Wrote a Python data-validation pipeline for 8 external vendor feeds, reducing data quality errors by 60% month-over-month. Delivered quarterly regulatory submissions for 4 jurisdictions on time for 2 consecutive years.