SiddharthChandgude

Machine Learning Engineer in Pune. I build language-model agents that finish the conversation, and the prediction models and data pipelines underneath them.

Email me See the work

70 of 100

Fig. 1 Insurance and member queries resolved from start to finish by the agent I built. The rest reach a person with the context attached. Earlier work: churn detection up 30%, investment targeting up 20%.

3.5 years of putting machine learning into production.

I work on generative AI, natural-language processing and analytics that people rely on at work: conversational applications built on large language models, agent workflows, retrieval-augmented generation and systems where several agents cooperate. Underneath sit Python, SQL, Azure, AWS, FastAPI, LangChain, AutoGen and vector databases. The results so far are in customer conversations handled without a person, operational analytics, investment targeting and churn prediction.

Three things I built, and what changed.

I was charged twice for my premium this month.

Heard as a billing problem: duplicate charge

I can see two debits on the 3rd. I have reversed the second one. It reaches your account in 3 to 5 working days.

Resolved without a person

Fig. 2 The kind of call it handles. An illustration, not a real transcript.

A virtual agent that finishes the call

An Azure-hosted language-model agent that answers insurance and member queries by itself. It identifies the topic while the caller is still talking, works out what they want, and escalates with context when a person is needed.

70% of queries resolved end to end

Azure · LLM · NLP · Transformers · FastAPI

SG Analytics, in production

Fig. 3 For every 10 departing clients caught in time before, the model catches 13. An illustration of the ratios, not client data.

Knowing which wealthy clients are about to leave

A prediction pipeline on financial and demographic data that scores how likely a high-net-worth client is to invest, and how likely to go. Feature engineering runs by itself and relationship managers get an alert in real time.

+30% churn detection, +20% targeting accuracy

XGBoost · scikit-learn · Python · SQL · AWS

Aion Tech Solutions, enterprise

Fig. 4 Loose operational data in, one readable chart out. An illustration.

Analytics pipelines that do the digging

Python and SQL pipelines feeding executive dashboards. They surface operational patterns, cut the manual analysis people used to do by hand, and back decisions about where to put resources.

Python · SQL · Pandas · Dashboards

SG Analytics, analytics

Where I have worked.

2023202420252026

SG Analytics2 yr 6 mo

Aion Tech Solutions Ltd.1 yr 1 mo

Machine Learning EngineerSG Analytics · Pune, India

April 2024 to now

Built an Azure-hosted language-model virtual agent for insurance and member queries. It resolves 70% of them from start to finish.
Wrote transformer-based workflows that pick out the topic of a call as it happens, work out what the caller wants, and hand over to a person with the context when needed.
Built Python and SQL analytics pipelines and executive dashboards that surface operational patterns and back resource-allocation decisions.

Machine Learning EngineerAion Tech Solutions Ltd. · India

March 2023 to March 2024

Built predictive models on financial and demographic data to find which clients are likely to invest and which are likely to leave.
Raised investment-targeting accuracy by 20% and churn detection by 30%.
Automated feature engineering and real-time alerts used to retain high-net-worth clients.

What I work with.

Agentic and generative AI: Agentic AI, AutoGen, LangChain, LlamaIndex, RAG, Prompt engineering, Function calling, Multi-agent systems, Memory design, Guardrails
Language models and NLP: GPT-4, GPT-3.5, BERT, Hugging Face, Semantic search, Embeddings, Sentiment analysis, NER, Topic modelling
Machine and deep learning: PyTorch, TensorFlow, scikit-learn, XGBoost, LightGBM, CNNs, Transfer learning, Time series
Data and engineering: Python, SQL, Pandas, NumPy, Apache Spark, FastAPI, Flask, REST APIs, Git
MLOps and cloud: Azure, AWS, SageMaker, Docker, Kubernetes, Airflow, CI/CD, Pinecone, FAISS
Analytics: Exploratory analysis, A/B testing, Hypothesis testing, Statistical modelling, Matplotlib, Seaborn, Dashboards

Certificates and degree.

Machine Learning in ProductionCertificate · DeepLearning.AI
Advanced Data Analytics CertificateCertificate · Google
B.E. Computer Science (AI, ML, Data Science)Degree · Pune University · 71.33%

Have a role or a project? Write to me.

siddchandgude@gmail.com

+91 8669053722Pune, Maharashtra, India

I am looking for machine learning and generative AI engineering roles, and I take on consulting work on language-model agents and ML pipelines.