Sravan Babu Bodapati

Amazon (United States) · United States

Unclaimed · AI-structured profileOpenAlex

Amazon researcher bridging Text-to-SQL multi-agent RL and efficient long-context LLM inference, with speech/NLP roots.

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-16) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.

Research profile

Output—

Output data is insufficient: the input provides no total work count, authorship-position distribution, or venue list, so this axis cannot be scored yet. It would measure cumulative publication volume and lead-authorship share across the full record.

Impact—

Impact data is unavailable: no per-paper citation counts, citation distribution, or venue-tier metrics are provided, so this axis is not scored. It would measure the citation footprint and scholarly uptake of the researcher's publications over time.

Focus0/100

The focus score is 0, consistent with the raw field distribution: recent reviewed work splits evenly (0.5/0.5) between Text-to-SQL natural language processing and efficient long-context LLM inference, while OpenAlex topics span topic modeling (39), speech recognition and synthesis (34), NLP techniques (28), and speech and audio processing (18 and 13). The record is currently broad rather than concentrated in a single niche.

Collaboration55/100

Collaboration scores 55. OpenAlex logs 15 distinct coauthors over the last three years, and 12 recurring collaborators are listed, including Shijie Chen, Yeounoh Chung, Yu Su, and Fatma Ozcan. On both reviewed 2026 papers Bodapati is a middle author and not the corresponding author, indicating work inside larger teams rather than solo-led efforts.

Activity90/100

Activity is the strongest axis (score 90). OpenAlex records 21 works in the last three years, with 17 publications in 2023, 13 in 2025, and 2 already in 2026 against a baseline of 2 in 2018 — a sustained and rising cadence. Both Evolve-reviewed papers appeared in 2026, confirming that current output continues at pace.

Reproducibility55/100

Reproducibility scores 55, computed from Evolve's reviews of 2 papers (mean readiness 55). Both are rated medium: DualSQL's ablations are solid but its margin over 32B baselines is thin, and SGD-KV shows train/eval overlap with no variance statistics tempering its SOTA claim. Results look checkable, though independent replication would still require care.

Research summary

Sravan Babu Bodapati is a US-based researcher at Amazon whose profile is name-resolved from public OpenAlex bibliographic data (no ORCID is linked). The record shows an h-index of 12 and an i10-index of 13, with 21 works in the last three years and 15 distinct coauthors; annual counts rise from 2 (2018) to a 2023 peak of 17, then 6 (2024), 13 (2025), and 2 already in 2026. OpenAlex topics are dominated by topic modeling (39), speech recognition and synthesis (34), NLP techniques (28), speech and audio processing (18), and speech and dialogue systems (13), indicating a base in speech and language technology. The two papers Evolve has reviewed, both 2026 arXiv preprints with middle authorship, match the current field split: DualSQL applies multi-agent reinforcement learning on a Qwen3 backbone to Text-to-SQL (71.1% execution accuracy on BIRD-Dev), and SGD-KV uses summarization-head scores to compress the KV cache by 75% while holding up on 1M-token contexts. Evolve judges both technically solid with caveats — a thin margin over 32B baselines in one case, train/eval overlap and missing variance statistics in the other — for a mean reproducibility score of 55 across 2 papers. Output and impact axes remain unscored for lack of data.

Collaboration pattern

Bodapati works in sizeable teams: OpenAlex counts 15 distinct coauthors over the last three years, and the catalog lists 12 frequent collaborators, including Shijie Chen, Yeounoh Chung, Yu Su, Fatma Ozcan, Yu Gan, and Jiani Zhang. On both reviewed 2026 papers he is a middle author and not the corresponding author, consistent with a contributor role within larger teams rather than solo-led projects; the collaboration dimension scores 55.

Research keywords

Text-to-SQLMulti-agent reinforcement learningSchema linkingGRPO reward designExecution accuracyKV cache compressionLong-context LLM inferenceAttention head specializationTopic modelingSpeech recognition and synthesisSpeech and dialogue systemsNatural language processing techniques

Highlights

DualSQL: multi-agent RL reaches 71.1% execution accuracy on BIRD-Dev
Joint multi-agent reinforcement learning on a single Qwen3 backbone lifts Text-to-SQL to 71.1% EX on BIRD-Dev, combining schema linking, GRPO-based training, and reward design. Evolve rates the ablations solid but notes the margin over 32B baselines is thin, with medium reproducibility readiness.
SGD-KV cuts KV cache by 75% using summarization-head scores
Allocating the KV budget by summarization-head scores compresses the cache by 75% while delivering solid results on 1M-token contexts, exploiting attention head specialization. Evolve flags train/eval overlap and the absence of variance statistics as caveats on the SOTA claim, with medium reproducibility readiness.
Two 2026 preprints span agentic NLP and inference efficiency
Within the same year, the Evolve catalog pairs a Text-to-SQL multi-agent RL system with a KV-cache compression method, mirroring the even 0.5/0.5 split between Text-to-SQL and efficient long-context LLM inference in the researcher's field distribution.

Timeline

2018Earliest year in the available OpenAlex counts-by-year window, with 2 publications recorded.
2021Strongest year of the early period with 9 publications, up from 4 in 2020 and 6 in 2019.
2023Peak output year in the record, with 17 publications, up from 8 in 2022.
2025Output rebounds to 13 publications after 6 in 2024, keeping the recent three-year total at 21 works.
2026Two papers enter the Evolve catalog: DualSQL (Text-to-SQL multi-agent RL) and SGD-KV (KV cache compression); 2 publications are already counted for the year.

Papers in the Evolve catalog

2026-09-17
DualSQL: Text-to-SQL with Multi-Agent Reinforcement Learning
2026-09-03
SGD-KV: Summarization Guided KV Cache Compression

Frequent co-authors

Shijie Chen×1 Yu Gan×1 Yeounoh Chung×1 Jiani Zhang×1 Quannan Li×1 Cody J. Greer×1 Yu Su×1 Fatma Ozcan×1 Zeyu Liu×1 Woomin Song×1 Xuandi Fu×1 Sai Muralidhar Jayanthi×1

About this profile

This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.

Bibliometrics are from OpenAlex only; no ORCID is linked, so this profile is name-resolved from public bibliographic data and identity cannot be independently confirmed.
Output and impact dimensions are reported as missing because no total-work counts, per-paper citations, or venue metrics were provided; h-index (12) and i10-index (13) are the only citation-derived figures available.
Only two papers have Evolve reviews, both 2026 arXiv preprints with middle authorship, so the recent field split (Text-to-SQL vs. efficient long-context inference) rests on a small, preprint-heavy sample; peer-review status and final venues are not reflected.
OpenAlex topics include a substantial speech and audio cluster (speech recognition and synthesis 34, speech and audio processing 18) that is not covered by the Evolve-reviewed papers; career events cannot be dated beyond the provided publication counts.