PIYUSH NIKAM

AI Researcher

Bengaluru, Karnataka, India9 yrs 9 mos experience

Most Likely To SwitchAI Enabled

Key Highlights

Expert in machine learning and anomaly detection.
Led impactful projects in SaaS and e-commerce.
Published research in top-tier conferences.

Stackforce AI infers this person is a Data Scientist specializing in SaaS and e-commerce with a strong focus on machine learning.

Contact

Skills

Core Skills

Machine LearningDeep LearningLarge Language Models (llm)NlpAnomaly DetectionComputer Vision

Other Skills

AlgorithmsAnalytical SkillsApache SparkApplied Machine LearningArtificial Intelligence (AI)Business RequirementsBusiness UnderstandingCC (Programming Language)C++CommunicationConvolutional Neural Networks (CNN)Data AnalysisData MiningData Modeling

About

Lead Data Scientist with 8+ years of experience across SaaS, e-commerce, and R&D, leveraging machine learning to build real-world, scalable systems. At Freshworks, I lead initiatives combining classical ML and LLMs to deliver interpretable, impact-focused solutions for enterprise customers. I specialize in fraud and anomaly detection, reinforcement learning, and ML optimization, and actively integrate prompt-tuned LLMs into RCA (Root Cause Analysis) and summarization workflows using tools like DSPy. Previously at Swiggy, I developed an unsupervised delivery executive anomaly ranking system using deep learning and tree-based ML models, influencing critical workflows like order assignment and platform actioning. At Qualcomm, I focused on optimizing transformer-based NLP models on Cloud AI 100, improving inference power efficiency by 52% through mixed-precision quantization and model compression. I hold a full-time Master’s in Computer Science from BITS Pilani, with published research in federated learning and adversarial ML using reinforcement learning. I’m deeply passionate about solving complex problems at scale, designing robust and explainable models, and bridging the gap between research and production. Always curious, always building. I enjoy connecting with interesting people and learning new things. Feel free to reach out. I look forward to hearing from you!

Experience

9 yrs 9 mos

Total Experience

1 yr 11 mos

Average Tenure

2 yrs 6 mos

Current Experience

Freshworks

Lead Data Scientist

Dec 2023 – Present · 2 yrs 6 mos · Bengaluru, Karnataka, India · Hybrid

Designed and deployed a hybrid ML+LLM system that integrates DSPy-powered prompt engineering with classical machine learning models to generate real-time, explainable insights. This system powers multiple analytics use cases such as RCA (Root Cause Analysis) and LLM-as-a-judge workflows, where model predictions are validated or interpreted via curated LLM reasoning. The architecture balances interpretability, latency, and business impact while leveraging prompt modularity and response templates.
Built an AI-powered Chart Summarization engine that parses visual elements (e.g., bar, line, pie, combo charts) and generates contextual narrative summaries and key insights from user dashboards in real time. Combined structured data parsing, NLP techniques, and prompt chaining to enable users to consume analytical stories without manual interpretation, drastically improving decision-making efficiency for enterprise users.

Python (Programming Language)Deep LearningForecastingStatistical ModelingAnomaly DetectionLarge Language Models (LLM)+1

Swiggy

Data Scientist 2

Jan 2022 – Dec 2023 · 1 yr 11 mos · Bengaluru, Karnataka, India · Remote

Trust & Safety – Swiggy (Food & Instamart Platforms)
Led the design and deployment of fraud and anomaly detection models for both delivery partners (DE) and customers, using a combination of unsupervised and semisupervised methods to capture complex behavioral patterns.
Developed the Delivery Executive Anomaly Model, a scalable fraud scoring system leveraging multi-window behavioral features, zone-normalization, and ensemble voting, powering workflows like order assignment, incentives, and actioning.
Collaborated cross-functionally with finance team, PMs, risk management (RMT), delivery platform, and engineering teams to drive alignment on model thresholds, policy decisions, and platform-level actioning.
Adapted to city-tier and seasonal drift (e.g., festivals, rain, IPL spikes) via zone-aware tuning and real-time monitoring workflows in Databricks and Snowflake, ensuring stable precision-recall over time.
Integrated the fraud detection outputs into key business levers like order assignment, DE incentivization, and compliance auditing, significantly reducing Swiggy-Driven Cancellations (SDC) and increasing platform integrity.

Python (Programming Language)Business RequirementsDeep LearningExploratory Data AnalysisApplied Machine LearningBusiness Understanding+2

Netapp

2 roles

Member Technical Staff 2

Jan 2020 – Jan 2020 · 0 mo · Bengaluru, Karnataka, India · On-site

Built a defect prediction system during BITS Practice School, leveraging project metadata and historical commit information to proactively identify modules at higher risk of critical defects, aiding in targeted QA and resource allocation.
Contributed to SnapMirror, NetApp’s data replication technology within the ONTAP team, focusing on cross-platform data protection and simplifying backup and disaster recovery workflows across the Data Fabric.

Business RequirementsExploratory Data Analysis

Member Technical Staff Intern

May 2019 – Jul 2019 · 2 mos · Greater Bengaluru Area · On-site

Contributed to the development of SnapMirror within the ONTAP team, enabling cross-platform replication and simplifying data protection across the Data Fabric.
Received a Pre-Placement Offer (PPO) for exceptional performance and contributions to enterprise-grade data replication systems.

Business Requirements

Qualcomm

Machine Learning Engineer

Jan 2020 – Jan 2022 · 2 yrs · Bengaluru, Karnataka, India · Hybrid

Spearheaded optimization of computer vision and NLP models for edge devices and cloud inference using quantization, mixed-precision techniques, and model compression, achieving up to 86% performance improvement and 52% better power efficiency on the Qualcomm® Cloud AI 100 platform.
Engineered scalable ML solutions aligned with the AI inference requirements of data centers, automotive systems, and edge deployments, addressing stringent latency, power, and memory constraints.
Benchmarked and accelerated transformer-based NLP models, deploying them on Qualcomm AI 100 using optimized runtime libraries and hardware-aware tuning.
Played a key role in advancing the ADAS (Advanced Driver Assistance Systems) pipeline, optimizing model throughput to support real-time decision-making.

Python (Programming Language)Deep LearningQuantizationMachine LearningComputer Vision

Birla institute of technology and science, pilani - goa campus

Teaching Assistant

Aug 2018 – Dec 2019 · 1 yr 4 mos · Goa, India · On-site

Conducted research in adversarial machine learning, reinforcement learning, and federated learning as part of a full-time Master's program at BITS Pilani. Focused on designing robust ML systems under adversarial settings and limited supervision.
Published multiple papers in top-tier conferences, including IJCNN, MILCOM, CODS-COMAD, and Springer LNCS, covering topics such as: Adversarial attacks and defenses in Android malware detection using reinforcement learning, Federated teacher-student self-training for heterogeneous UAVs, Fraud detection in food delivery platforms using semi-supervised deep anomaly detection (DevNet + variational loss)
Led end-to-end experimentation: from data preprocessing, model design (RL, federated, semi-supervised), training, to performance benchmarking and paper writing.
Also served as a Teaching Assistant for Data Mining and Artificial Intelligence courses, conducting ML-focused lab sessions and assisting with lecture content.

Python (Programming Language)Deep LearningMachine Learning

Manhattan associates

Software Engineer

Jan 2016 – Jan 2018 · 2 yrs · Bengaluru, Karnataka, India · On-site

Designed and built NLP-driven solutions to address real-world business challenges, leveraging frameworks like DialogFlow, Alexa Skills Kit, and custom NLU pipelines.
Developed and integrated a virtual assistant with the Order Management System to automate customer support workflows and reduce manual load.
Engineered a scalable, rule-based promotion management system as part of Manhattan Active™ Omni, enabling dynamic offer configurations and improving campaign agility.

Business RequirementsMachine Learning