Jun 2024 – Apr 2025 · 10 mos · New Delhi, Delhi, India · Remote

• 1. Facilitated 15+ technical deep-dive sessions and provided structured mentorship to over 100 contributors, accelerating their contributions to the Digital Management Platform (DMP) and key open source community projects.
• 2. Designed and launched 3+ community engagement campaigns, resulting in a 40% increase in active contributors and a 25% improvement in contributor retention over six months.
• 3. Led diversity and inclusion initiatives, increasing women’s participation in community programs by 60% year-over-year, and implemented inclusive onboarding practices that improved newcomer engagement by 30%.

Jul 2023 – Apr 2025 · 1 yr 9 mos · New Delhi, Delhi, India · Remote

• 1. Ingested 1,887 articles into a RAG database, then ran statistical analysis, trend mapping, and segmentation. Basically turned raw text into structured insights that could actually be queried.
• 2. Cut average query response times by 25% on the Milvus Vector Database — meaning users got answers faster, without sacrificing precision.
• 3. Executed Test-Driven Development of 65 NLP models, benchmarking them on accuracy and latency, aligned with product success metrics.
• 4. Designed dashboards using Power BI to communicate RAG metric performance (Semantic Similarity, Precision, Relevance,Correctness, Context Recall and precision.) for model evaluation and stakeholder reporting.

Jun 2024 – Present · 2 yrs · New Delhi, Delhi, India · Remote

• 1. Dug through 50K+ operational records in MySQL to surface incident frequency, distribution quirks, and regional hotspots — insights that helped leadership deploy resources where they actually mattered. Dealing with ambiguous problems statements and messy data is my game.
• 2. Took raw, messy spreadsheets and rebuilt them with Excel (PivotTables, Power Query) —> cutting manual reporting time , mockup creation to final dashboard time by 2 weeks on average and giving executives data they could finally trust.
• 3. Designed Power BI dashboards with threat heatmaps and zone-level views across 12+ regions —> dashboards that became the go-to playbook for identifying risks and saving costs annually in operational inefficiencies. Basically if there is a regional hotspot to look at, I found it!
• 4. Partnered with cross-functional teams to define product KPIs, lifecycle stages, and success metrics, then ran Agile sprints through JIRA — boosting sprint completion rates by 18% and reducing product rollbacks.

Jan 2023 – Feb 2023 · 1 mo

• The Harvard WECode (Women Engineers Code) Conference is organized by undergraduate women at Harvard University and is an initiative of Harvard Undergraduate Women in CS. WECode hosts the largest student-run women in tech conference in the world.

Oct 2022 – May 2024 · 1 yr 7 mos · Remote

• 1. Utilized Python,Networkx, Matplotlib to implement Greedy Coloring LCA Gallery Example in documentation.
• 2. Wrote 15 Docstring Examples for enhanced usability of documentation for Bidirectional Path Dense Algorithm
• 3. Executed the fixing of 5 hyperlinks to Eigenvector and katz centrality files totalling to 18 Major Merged PRs
• 4. Uncovered Major Bug in Semiconnected.py by rigorous Unit Testing leading to 20% improvement in program
• 5. Merged 7 Pull requests to enhance documentation by adding Raises warning, docstrings and fixing hyperlinks.

Jun 2022 – Aug 2022 · 2 mos · India

• It was a great opportunity for students to learn from and interact with Scientists at Amazon who have immense knowledge in their ML domain. The eight modules on ML topics that were covered:-
• · Supervised Learning
• · Deep Neural Networks
• · Dimensionality Reduction
• · Unsupervised Learning
• · Probabilistic Graphical Models
• · Sequential Learning
• · Causal Inference
• · Reinforcement Learning

Mar 2022 – Jul 2022 · 4 mos

• Filtered IIT Bombay English-Hindi Corpus using Wikipedia resulting in 30% reduction in noise irrelevant data
• ∗ Implemented techniques of tokenization and part-of-speech tagging, to effectively filter corpus based on Wikipedia.
• 2. Formulated Research Testing Procedure for mBART and BLOOM leading to 80% Test Coverage
• 3. Achieved a 86% in BLEU score and 89% in METEOR score after hyper-parameter tuning.
• 4. Implemented targeted modifications in model architectures, resulting in 12% increase in accuracy.

Mar 2022 – Jun 2022 · 3 mos · Delhi, India

Dec 2021 – Feb 2022 · 2 mos · India