Looking for a job

Praneet Sahgal

@praneetsahgal

Message

Senior Machine Learning Engineer specializing in production Generative AI systems.

United States

Message

What I'm looking for

I’m seeking a Lead Generative AI role to drive technical direction, mentor teams, and own the full AI product lifecycle—from architecture and evaluation through deployment and continuous improvement.

I’m a Senior Machine Learning Engineer with 10+ years building production AI systems across generative AI, deep learning, computer vision, and edge inference. I own high-ambiguity problems end to end, and I’m known for raising engineering standards across teams by building reliable systems, not just models.

I’ve built end-to-end GenAI platforms spanning the full stack—RAG architecture, agentic workflows, SLM/LLM fine-tuning, and LLMOps observability. At Adobe, I architected an agentic AI platform using a LangGraph-based orchestration engine with strict schema-validated function calling and policy guardrails, deployed a multi-model inference stack, and established LangSmith tracing to cut agent failure root-cause analysis time from hours to minutes. Earlier roles strengthened my production instincts: I built prediction and forecasting systems at Peloton (Temporal Fusion Transformer, LSTM/GRU/XGBoost ensembles), developed real-time 3D pose estimation pipelines (PyTorch, quantization, pruning, TensorRT), and delivered edge-optimized vision models (TensorFlow to TensorFlow Lite, INT8 quantization) that met tight latency and accuracy constraints.

Experience

Work history, roles, and key accomplishments

Current

Senior Machine Learning Engineer

Current

Adobe

Aug 2024 - Present (1 year 11 months)

Architected Adobe’s internal agentic AI platform with LangGraph-based orchestration, dynamic model routing, and schema-validated tool calling with policy guardrails. Built production RAG and an NL→SQL SLM fine-tuning workflow using QLoRA/LoRA with LLM evaluation and LLMOps observability in LangSmith.

LangGraph Agentic UI Flows Schema Validated Function Calling VLLM Langsmith Bi Encoder BM25 LLM As Judge Prompt Versioning Drift Monitoring QLoRA PEFT Instruction Tuning RBAC Audit Logging

Senior Machine Learning Engineer

Peloton Interactive

Aug 2022 - Aug 2024 (2 years)

Built a behavioral analytics platform for churn and workout-frequency prediction, including a Temporal Fusion Transformer and ensemble forecasting models. Developed a multi-turn LLM fitness coaching assistant with RAG grounding and production ML pipelines for evaluation, A/B testing, and drift monitoring.

Temporal Fusion Transformers XGBoost Anomaly Detection DCGAN Synthetic Data Generation LangGraph RAG VLLM TorchServe Feast Shadow Deployments A B Testing Drift Monitoring Evidently AI WhyLabs

Senior Computer Vision Engineer

Peloton Interactive

Aug 2020 - Aug 2022 (2 years)

Built Peloton Guide’s real-time 3D body pose estimation pipeline using PyTorch keypoint lifting from 2D RGB. Developed temporal pose scoring with GRU models and optimized the inference stack using INT8 quantization, pruning, and TensorRT for real-time performance.

PyTorch Pose Estimation 2D To 3D Transformation GRU Temporal Pose Model Pruning TensorRT Synthetic Data Generation Rep Counting Signal Generation Form Scoring

Computer Vision Scientist

Affectiva

Apr 2019 - Jun 2020 (1 year 2 months)

Ported face detection and facial landmark models from TensorFlow to TensorFlow Lite for automotive SoCs, optimizing for size and latency while maintaining accuracy. Developed child presence and forgotten object detection models with hard real-time constraints using hardware-in-the-loop benchmarking.

TensorFlow Tensorflow Lite Facial Landmark Models Child Monitoring Forgotten Object Detection

Deep Learning Engineer

Neurala

Feb 2016 - Apr 2019 (3 years 2 months)

Built an AI-assisted video annotation module using model-in-the-loop active learning in Brain Builder to reduce annotation effort and labeled data needs. Developed 3D vision and motion-capture systems, exporting optimized inference via ONNX/TensorRT for deployment on Intel Movidius NCS and Qualcomm 845.